test: build the purge-response fixture ARN at runtime

The rebrand compatibility guard tracks every literal policy value that names the upstream brand. Construct the replication ARN the way the other replication fixtures do instead of adding a new literal. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Feng Ruohang <rh@vonng.com>
docs: record the delete-marker purge, listing quorum and migration tag repairs
2026-09-30 23:05:59 +03:00 · 2026-09-16 23:50:09 +08:00 · 2026-09-16 23:44:44 +08:00 · 2026-09-16 22:53:12 +08:00 · 2026-09-16 22:35:08 +08:00 · 2026-09-16 22:24:27 +08:00
340 changed files with 7648 additions and 25807 deletions
@@ -90,6 +90,18 @@ jobs:
      - name: Run S3 Select tests under race detector
        run: go test -race ./internal/s3select/... -count=1

+      - name: Run conditional PUT tests under race detector
+        run: go test -race ./cmd -run '^Test(PoolsConditionalPut|SinglePoolConditionalPutHTTP)' -count=1 -timeout=5m
+
+      - name: Run multipart listing and cancellation tests under race detector
+        run: go test -race ./cmd -run '^Test(MultipartListing|MultipartAbort|PaginateMultipartUploads|ListMultipartUploads)' -count=1 -timeout=5m
+
+      - name: Run CPU metrics tests under race detector
+        run: go test -race ./cmd -run '^TestLoadCPUMetrics' -count=1 -timeout=5m
+
+      - name: Run tag replication tests under race detector
+        run: go test -race ./cmd -run '^TestAPITagging' -count=1 -timeout=5m
+
  crosscompile:
    name: Cross Compile
    runs-on: ubuntu-latest
@@ -132,9 +132,9 @@ jobs:
          cosign-release: v3.1.2

      - name: Build Draft release with GoReleaser
-        uses: goreleaser/goreleaser-action@v7
+        uses: goreleaser/goreleaser-action@f06c13b6b1a9625abc9e6e439d9c05a8f2190e94 # v7.2.3
        with:
-          version: "~> v2"
+          version: v2.18.1
          args: release --clean --skip=validate --config .github/goreleaser.yml
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -0,0 +1,100 @@
+name: Repository Cards
+
+on:
+  schedule:
+    # 00:00 UTC = 08:00 Asia/Shanghai. GitHub may queue scheduled runs.
+    - cron: "0 0 * * *"
+  workflow_dispatch:
+  push:
+    branches: [main]
+    paths:
+      - .github/workflows/repository-cards.yml
+      - .github/silo.svg
+      - buildscripts/repository-cards/**
+  pull_request:
+    branches: [main]
+    paths:
+      - .github/workflows/repository-cards.yml
+      - .github/silo.svg
+      - buildscripts/repository-cards/**
+
+permissions:
+  contents: read
+
+concurrency:
+  group: repository-cards-${{ github.event.pull_request.number || 'publish' }}
+  cancel-in-progress: false
+
+jobs:
+  check:
+    name: Validate repository cards
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+        with:
+          persist-credentials: false
+      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+        with:
+          python-version: "3.13"
+      - run: python -m pip install -r buildscripts/repository-cards/requirements.txt
+      - run: python -m unittest discover -s buildscripts/repository-cards -p 'test_*.py' -v
+
+  publish:
+    name: Update README images
+    needs: check
+    if: github.repository == 'pgsty/silo' && github.ref == 'refs/heads/main' && github.event_name != 'pull_request'
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    permissions:
+      contents: write
+      issues: read
+      pull-requests: read
+    env:
+      OUTPUT_BRANCH: codex/repository-cards
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+        with:
+          python-version: "3.13"
+      - run: python -m pip install -r buildscripts/repository-cards/requirements.txt
+
+      - name: Load the generated-assets branch
+        shell: bash
+        run: |
+          set -euo pipefail
+          artifacts="$RUNNER_TEMP/repository-cards"
+          if git ls-remote --exit-code --heads origin "$OUTPUT_BRANCH"; then
+            git fetch --depth=1 origin "$OUTPUT_BRANCH"
+            git worktree add --detach "$artifacts" FETCH_HEAD
+          else
+            status=$?
+            # Exit 2 means no matching ref; transport/auth failures must stop.
+            if [ "$status" -ne 2 ]; then exit "$status"; fi
+            git worktree add --detach "$artifacts" HEAD
+            git -C "$artifacts" checkout --orphan "$OUTPUT_BRANCH"
+            git -C "$artifacts" rm -rf .
+          fi
+
+      - name: Refresh contributor and star cards
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: python buildscripts/repository-cards/update.py --output "$RUNNER_TEMP/repository-cards"
+
+      - name: Publish changed assets
+        shell: bash
+        run: |
+          set -euo pipefail
+          cd "$RUNNER_TEMP/repository-cards"
+          git add README.md history.json curated.json contributors.json \
+            contributors-light.svg contributors-dark.svg \
+            star-history-light.svg star-history-dark.svg
+          if git diff --cached --quiet; then
+            echo "Repository cards are already current."
+            exit 0
+          fi
+          git config user.name 'github-actions[bot]'
+          git config user.email '41898282+github-actions[bot]@users.noreply.github.com'
+          git commit -s -m "chore: update repository cards $(date -u +%F)"
+          # A normal push preserves history and refuses concurrent overwrites.
+          git push origin "HEAD:refs/heads/$OUTPUT_BRANCH"
@@ -4,6 +4,8 @@ on:
  workflow_dispatch:
  pull_request:
    paths:
+      - "go.mod"
+      - "go.sum"
      - ".github/goreleaser.yml"
      - ".github/nfpm.yml"
      - "Dockerfile.goreleaser"
@@ -93,9 +95,9 @@ jobs:
          echo "LDFLAGS: ${LDFLAGS}"

      - name: GoReleaser config check
-        uses: goreleaser/goreleaser-action@v7
+        uses: goreleaser/goreleaser-action@f06c13b6b1a9625abc9e6e439d9c05a8f2190e94 # v7.2.3
        with:
-          version: "~> v2"
+          version: v2.18.1
          args: check --config .github/goreleaser.yml

      - name: Validate Helm chart and legacy upgrade identity
@@ -107,9 +109,9 @@ jobs:
          syft-version: v1.50.0

      - name: Build snapshot artifacts
-        uses: goreleaser/goreleaser-action@v7
+        uses: goreleaser/goreleaser-action@f06c13b6b1a9625abc9e6e439d9c05a8f2190e94 # v7.2.3
        with:
-          version: "~> v2"
+          version: v2.18.1
          # A pull-request snapshot has no trusted release identity. Exercise
          # the SBOM/checksum pipeline here, and reserve keyless signing for
          # the tag-triggered release workflow with GitHub OIDC.
@@ -62,10 +62,10 @@ dist/

 .claude/
 .codex/
-AGENTS.md
-CLAUDE.md
 _bmad/
 _bmad-output/
+# Investigation and editorial working files belong outside this repository.
+docs/investigations/
 docs/security/
 docs/rebranding.md
 .release-sign/
@@ -0,0 +1,84 @@
+# SILO Repository Guide
+
+## Project map
+
+- Server: `pgsty/silo` (this checkout, default branch `main`)
+- Console: `pgsty/silo-console` (usual sibling checkout `../silo-console`)
+- Client: `pgsty/mc`, shipped as `mcli` (usual sibling checkout `../mc`)
+- Shared packages: `pgsty/silo-pkg` (usual sibling checkout `../silo-pkg`)
+- Documentation: `pgsty/silo.pgsty.com` (usual sibling checkout `../silo.pgsty.com`),
+  published at <https://silo.pgsty.com/>
+
+The server executable, package, systemd service, and container image use `silo`;
+the public Docker image is `docker.io/pgsty/silo`. The server module remains
+`github.com/minio/minio`. Consult the current `go.mod` and release configuration
+for selected component versions and compatibility replacements.
+
+## Supported stack and compatibility policy
+
+The maintained, release-gating product graph is the coordinated PGSTY stack:
+`silo` + `silo-console` + `mc` + `silo-pkg`.
+
+Keep compatibility with upstream MinIO/MC on a best-effort basis. Preserve
+inexpensive wire, configuration, CLI, migration, and import compatibility when
+it helps users, and document known differences. Do not infer from source
+lineage, MinIO-compatible protocols, retained `MINIO_*`/`MC_*` names, or an
+inherited upstream test that unmodified upstream MinIO is a supported release
+target.
+
+Upstream-only compatibility checks may remain as advisory evidence, but they
+must not force a downgrade of a maintained SILO component, a fork-only API
+workaround, or a release block. Make upstream compatibility a hard gate only
+when the user explicitly requests that scope.
+
+## Dependency selection
+
+- When PGSTY maintains a component, import and require it directly under its own
+  module path. In particular, prefer `github.com/pgsty/silo-pkg/v3` over
+  `github.com/minio/pkg/v3` in maintained SILO source.
+- Do not use an upstream package merely to make unmodified upstream MinIO or MC
+  compile. Unavoidable transitive upstream modules should be documented and
+  kept separate from the maintained product dependency.
+- `github.com/minio/minio-go/v7` is the explicit exception: use the verified
+  upstream module/commit while it contains the required fixes; do not recreate
+  a SILO fork without a concrete functional divergence.
+- Check the current `go.mod` before changing versions. Coordinate breaking
+  import-path changes across package, client, Console, and server releases.
+
+## Documentation and delivery
+
+The companion site owns product, operations, migration, security-advisory,
+design, and release documentation. Maintain English and Chinese content in
+that repository and link to its canonical URLs from this one. The site's
+`content/docs/` is the documentation entry point; detailed pages live under
+`content/operations/`, `administration/`, `reference/`, `compatibility/`,
+`about/`, and `blog/`, following the existing structure.
+
+Keep repository entry points and contributor instructions here. Retained
+upstream documents, examples, and test fixtures under `docs/` may be updated
+when a code or tooling change requires it; do not build a second documentation
+site in this tree.
+
+Investigation plans, prompts, AI session transcripts, execution logs, temporary
+reports, and private security material belong outside the checkout, for example
+in a task-specific directory under `~/tmp/`. Do not recreate
+`docs/investigations/`, `docs/security/`, or `docs/rebranding.md`, or move their
+working material to another repository directory. Extract reusable, verified
+knowledge into the companion site; keep private evidence outside public Git.
+
+Compatibility pages may continue to describe similarities with upstream, but
+must label that compatibility as best effort. The supported and tested server
+for Console and mcli administration is `pgsty/silo`.
+
+Before removing or moving documentation, check repository links and scripts,
+the companion site's links and anchors, and references from this guide. Use
+fixed commit URLs for historical evidence that should survive deletion from
+the current tree. Run the site's `make check` for site changes and this
+repository's `make rebrand-guard` for documentation cleanup that changes the
+identifier inventory. When new public URLs are involved, publish the site
+before publishing source changes that depend on those URLs.
+
+Keep changes in the repository that owns the affected surface. Treat every
+repository commit, tag, release, image, documentation update, and deployment as
+a separate deliverable. Follow `CONTRIBUTING.md`, including DCO sign-off
+(`git commit -s`). `CLAUDE.md` imports this guide so both agents use one policy.
@@ -2,16 +2,48 @@

 ## Unreleased

-The entries below describe source changes on main since the latest published Server.
+Preparation target: `RELEASE.2026-09-16T00-00-00Z` (package version
+`20260916000000.0.0`). The entries below describe the candidate changes since
+the latest published Server.
 **The latest published Server remains 20260903.** These changes are not in its
 binaries, packages or images. See the [component matrix](https://silo.pgsty.com/compatibility/versions/)
 and [complete commit range](https://github.com/pgsty/silo/compare/RELEASE.2026-09-03T13-18-01Z...main).

 ### Authorization and security

+- Synchronize CPU metrics reads with resource-metrics updates (#210), preventing
+  concurrent map access from terminating the server during Prometheus scraping.
+  Metric names, values and authentication requirements are unchanged.
+- Restrict embedded Console's anonymous sharing proxy to object-content GETs
+  at the configured S3 origin, and reject every redirect. Internal metrics,
+  system paths and non-download S3 operations cannot be reached through it.
+  Normal public, presigned and versioned downloads remain available without a
+  new setting; a full sharing-disable switch is not introduced. See
+  [Console #56](https://github.com/pgsty/silo-console/pull/56) and the
+  [design record](https://github.com/pgsty/silo-console/issues/52).
+  Thanks to Jiri Pejchal (@jiri-pejchal) for the report.
+- Persist IAM deletion revisions and parent revocation boundaries so stale site
+  events cannot restore deleted identities, policies or their older grants
+  (#191, #192). Peer deletion notifications reload committed storage; deliberate
+  recreation requires a newer revision, and credentials issued before the
+  parent's revocation remain invalid.
+  **Coordinated upgrade required:** upgrade every participating node and site.
+  Mixed old/new nodes sharing an IAM backend and rolling downgrade are
+  unsupported. Back up complete IAM storage and encryption material; an admin
+  export of live records omits deletion history. Reissue credentials for
+  recreated parents and explicitly reconcile pre-upgrade revocations whose
+  history is already lost. Restoring an older backup can lose later revocations;
+  keep affected sites isolated until reconciliation/rekeying is complete. See
+  [the operator runbook](https://silo.pgsty.com/operations/replication/iam-upgrade/).
+- Enforce an absolute HTTP/1 request-header deadline through the connection
+  wrapper (#196). Repeated small reads no longer extend that deadline, and
+  `--read-header-timeout` / `MINIO_READ_HEADER_TIMEOUT` now reaches the HTTP
+  server. HTTP/1 request bodies retain the rolling idle timeout; this does not
+  impose a total upload/download duration. A shorter setting also constrains
+  TLS handshake reads. The wrapper's strict header mode is not applied to HTTP/2.
 - Reject unsigned `x-amz-*` request headers that could turn a signed PUT into a
  copy of another object accessible to the signer (SN-2026-011). The latest
-  public Server is affected; the fix is on main. See [the advisory ledger](docs/security/advisories.md).
+  public Server is affected; the fix is on main. See [the advisory ledger](https://silo.pgsty.com/about/security-advisories/).
 - Align signed request fields with policy conditions and enforce header-only
  presigned payload checksums. See [the signed-header review](https://silo.pgsty.com/blog/design/signed-header-coverage/).
 - **Breaking policy semantics:** separate self-service `admin:ChangeMyPassword`
@@ -22,17 +54,114 @@ and [complete commit range](https://github.com/pgsty/silo/compare/RELEASE.2026-0

 ### Object storage and replication

+- Make `ListMultipartUploads` discover quorum-valid uploads from durable state
+  across pools, erasure sets and drives, then apply S3 prefix, delimiter,
+  marker, ordering and 1,000-entry pagination semantics globally (#198). New uploads
+  store their canonical bucket and key as reserved fields in the existing
+  quorum-written `xl.meta`; completion removes those upload-only fields. Native
+  markers remain usable after their upload is completed or canceled. Strict
+  listing returns a diagnostic 503 for legacy uploads or uncertain coverage;
+  the default remains the released exact-key/cache-based `legacy` behavior.
+  Opt into strict mode only through `MINIO_API_MULTIPART_LISTING=strict`, after
+  upgrading every writer, draining old uploads, checking the read-only admin
+  `multipart-preflight` report and validating scan capacity. The process-only
+  setting is not persisted into shared API configuration. Historical
+  `multipart_listing` keys are ignored and can be removed with a targeted
+  `mcli admin config reset ALIAS api multipart_listing` before rollback. Per-process
+  admission, directory-entry, worker and time budgets bound scan scheduling;
+  each page still scans durable state. See [issue #79](https://github.com/pgsty/silo/issues/79)
+  and its [design record](https://silo.pgsty.com/blog/design/list-multipart-uploads/).
+  Thanks to mr javad seydi (@mrjavadseydi) for the original implementation.
+- Retain released read-quorum and best-effort multipart cancellation in default
+  legacy mode. Strict mode requires majority deletion acknowledgements and
+  permits retries below read quorum. When most drives were already empty,
+  failed deletion of an observed remnant now returns 503 instead of being
+  masked by empty-drive successes. Wrong-key or wrong-bucket cancellation
+  preserves the valid upload's cache entry; successful cancellation notifies
+  peers with the request context after the distributed lock is released.
+  The HTTP response for an absent
+  upload remains 204; this is not proof of physical cleanup. **Known boundary:**
+  delayed creation writes can still restore an upload after cancellation;
+  this change does not add a durable creation fence.
+- Preserve object tags during multi-pool metadata reconciliation by reading the
+  resolved tag field together with its revision (#189). Previously, reconciliation
+  could replace existing tags with an empty value.
+- Preserve the tag revision on SSE-KMS metadata replication (#193), and advance
+  tag revisions monotonically on local PUT/DELETE tagging (#196). Empty tags
+  participate in reconciliation as an ordered deletion, preventing older
+  events from restoring removed tags. SSE-C key rotation also retains the tag
+  revision. Malformed historical revisions can fail and retry; their missing
+  history is not reconstructed by the upgrade.
+- Complete delete-marker version purges and preserve their identity and retry
+  state through MRF recovery (#196). Recovery accepts a 405 marker response only
+  when its version, bucket, object name and modification time match the task.
+  Purge audit status is normalized from `COMPLETE` to `COMPLETED`.
+  Thanks to Julien Laurenceau (@julienlau) for the investigation and proposed
+  fix in #184 that helped shape this follow-up.
+- Restore only the six replication-specific metadata fields after ordinary
+  request metadata extraction (#194). This prevents transport-only `aws-chunked`
+  from being stored as Content-Encoding while preserving the signed-header
+  protections. Trusted Snowball entries no longer inherit the outer archive's
+  ordinary metadata. Thanks to Mikhail Khadarenka (@chodorenko) for the fix in #187.
+  **Existing data:** these repairs prevent new errors; they do not scan or rewrite
+  historical object metadata, recover lost tags or prove that old purge work has
+  converged. Follow the [read-only audit procedure](https://silo.pgsty.com/operations/replication/replica-metadata-audit/)
+  before planning any repair of stored state.
+
+- Make exact-version delete-marker purges converge in replicated buckets
+  (`eb4f5e5b3`, `254b19ac0`, `358ab38fb`). A purge no longer creates the
+  marker on drives that lacked it; a retried purge of a missing version is
+  acknowledged only when a write-quorum majority of drives report it absent;
+  purge results merge removed and reliably absent replies; healing a marker
+  preserves its stored replication and purge metadata; and a queued marker
+  creation is re-checked against the source under the replication lock before
+  it is sent, so a purge that already reached the targets is not undone by a
+  stale task from another frontend, a GET/LIST heal, the scanner or MRF.
+  Purging a data version whose earlier purge is still pending reports it as a
+  data version. **Known limitations:** creations already in flight or replayed
+  from another site, and minority marker copies left by a crash after a
+  majority-acknowledged purge, are tracked in #217. See
+  [the replication reliability record](https://silo.pgsty.com/blog/design/replication-reliability/).
+- Keep a null object version that has listing quorum when a newer minority of
+  drives sorts first (`8d06424b1`). The resolver recounts per header only when
+  the original selection lacks quorum, every non-empty drive stream holds
+  exactly one ordinary null version and all share the same erasure layout;
+  mixed histories keep their previous behavior. **Known limitation:** a
+  successful ListObjects can still omit readable keys during rolling restarts
+  with concurrent overwrites (#218). Do not run destination-deleting sync tools
+  against a listing taken during a rolling restart; list again once the
+  cluster is stable.
+- Carry object tags through rebalance and decommission for ordinary and
+  multipart writes (`fced86303`). Both migration entry points restore the tags
+  and their revision fields when rewriting the object in the destination pool.
+  Tags dropped by earlier migrations are not recovered; audit tag-dependent
+  lifecycle and policy rules for pools migrated with an older build.
+
 - Evaluate conditional multipart completion against the logical current object
  across all pools while holding the existing object lock. A stale `If-Match`
  can no longer replace newer data in another pool, and the current ETag is no
  longer rejected because the upload resides next to an older copy. Conditions
  are evaluated once; a current delete marker counts as an absent object.
-  **Availability change:** if metadata cannot be read from any pool, conditional
+  **Availability change:** if any pool's metadata cannot be read, conditional
  completion fails even when another pool can still serve GET/HEAD. This also
  applies when the unreadable pool may not hold the object: absence cannot be
  verified. Retry after the pool recovers. Unconditional completion and the
  single-pool path retain their existing behavior.

+- Evaluate ordinary multi-pool conditional PUT against the logical current
+  object across all pools, including draining pools, under the existing object
+  lock (#207). A stale destination copy no longer accepts a stale ETag or rejects
+  the current one; a current delete marker is treated as absence.
+  **Availability change:** if any pool's object metadata cannot be verified,
+  the condition fails even when GET can use another pool; read-quorum failures
+  return 503. Restore readability or heal before retrying. Unconditional PUT,
+  single-pool conditions and internal replication retain their existing behavior.
+  A public condition with a destination `versionId` compares the current object
+  while preserving the requested write version. This change does not retire
+  stale copies in other pools, undo historical accepted overwrites or provide
+  a new global clock-ordering guarantee. The multipart-completion repair in #190
+  neither introduced nor repaired this separate PUT defect.
+
 - Reconcile ordinary single-object version DELETE across all pools, including
  null versions, delete markers and unqualified directory-marker DELETE. This
  applies the deletion to every resolved pool copy under existing quorum
@@ -50,7 +179,7 @@ and [complete commit range](https://github.com/pgsty/silo/compare/RELEASE.2026-0
  its tracker, mover, scanner hooks, configuration, XML actions and metrics.
  Accept and ignore retired configuration/XML and preserve ordinary statistics
  when reading v9 caches. See [migration notes](docs/bucket/lifecycle/access-tiering-removal.md).
-  The [decision record](docs/investigations/access-tiering-revert.md) preserves
+  The [decision record](https://silo.pgsty.com/compatibility/access-tiering-removal/) preserves
  the feature's introduction, subsequent fixes, rollback scope and review history.
 - Preserve the independent multi-pool write, metadata, healing and conditional
  deletion fixes from PR #178, including shared remote-tier reference protection.
@@ -61,10 +190,10 @@ and [complete commit range](https://github.com/pgsty/silo/compare/RELEASE.2026-0
 - Repair federated CopyObject checksums, destination timestamps, reserved
  metadata, encrypted-object forwarding, legal hold and KMS context.
 - Make resync counters, target selection, cancellation and worker lifetimes
-  reflect actual work; complete delete-marker purges and report bounded MRF drops.
+  reflect actual work, and report bounded MRF drops.
 - Converge bucket metadata with deterministic source state, deletion tombstones,
  creation time recovery and diagnostics. The mixed-version export gate requires
-  coordinated upgrades before tombstones are exported. See [the #77 record](docs/investigations/issue-77-current.md).
+  coordinated upgrades before tombstones are exported. See [the #77 record](https://silo.pgsty.com/blog/design/bucket-metadata-convergence/).
 - Include per-bucket CORS in metadata export/import, close metadata publication
  and logger races, and report effective bucket quotas in metrics.

@@ -72,23 +201,28 @@ and [complete commit range](https://github.com/pgsty/silo/compare/RELEASE.2026-0

 - Restore embedded Console login over loopback TLS, trusted-proxy handling and
  all four WebSocket connection limits. Preserve Go TLS defaults across transports.
- Directly require `github.com/pgsty/silo-pkg/v3` v3.14.0; select Console
-  `v0.0.0-20260913015128-417559bb2c97` and MC
-  `v0.0.0-20260913012246-4f609a4da3bb` with explicit PGSTY replacements.
- Pin upstream minio-go `v7.3.1-0.20260910142817-60bd07042d49`; refresh Go x/*
-  modules and security fixes including bounded AMQP frame handling. Keep Go
-  1.27.1 and go-systemd v22.6.0's NetBSD compatibility replacement.
+- Directly require `github.com/pgsty/silo-pkg/v3` v3.14.1; select released Console
+  v2.4.1 (`v0.0.0-20260916075814-1360e26d976d`) and mcli 20260916
+  (`v0.0.0-20260916070421-e952aa78f10a`) with explicit PGSTY replacements.
+  The embedded frontend identifies itself as Console v2.4.1.
+- Pin upstream minio-go `v7.3.1-0.20260915093545-32e1f32cb176` to handle
+  CopyObject errors embedded in HTTP 200 responses. Update JWX to v3.3.0 for
+  JSON field-name escaping, strfmt to v0.27.2 for Go 1.27 hostname validation,
+  and LZ4 to v4.1.30 for frame-reader, partial-read and concurrency fixes.
+  Retain the earlier Go x/* and bounded AMQP frame updates, Go 1.27.1, and
+  go-systemd v22.6.0's NetBSD compatibility replacement.
 - Refresh container base digests and build static curl 8.22.0 from verified
-  source for both Linux architectures. Pin the actual mcli 20260913 archives and
-  hashes. Helm's client image follows that release; its Server image still names
-  the latest published Server 20260903.
+  source for both Linux architectures. Pin the published mcli 20260916 archives
+  and hashes in the container and update the client installer default.
+- Prepare Helm chart 7.0.3 with Server and client defaults for the September 16
+  batch. Publish the chart only after the corresponding Server image exists.
+- Pin GoReleaser v2.18.1 and its action commit identically in snapshot and release
+  workflows. Dependency-only PRs now run the Test Release Pipeline too.

-The dependency update passed the final candidate's Go, vulnerability and Test
-Release workflows; native curl builds passed on both architectures. A local
-ARM64 image passed startup, health, S3 transfer and embedded Console checks.
-These checks do not publish a Server tag or production image and do not replace
-cluster upgrade/rollback acceptance for the next release. Dated investigations
-retain the exact source and runtime boundaries they tested.
+Validation of earlier source revisions does not establish acceptance of this
+candidate. Final source, package, image and multi-process checks are tracked
+separately in [#203](https://github.com/pgsty/silo/issues/203). No Server release
+or production rollout is implied by this preparation target.

 ## RELEASE.2026-09-03T13-18-01Z

@@ -0,0 +1 @@
+@AGENTS.md
@@ -77,6 +77,16 @@ test evidence, compatibility notes, and documentation impact. Public product
 documentation is owned by the separate
 [`pgsty/silo.pgsty.com`](https://github.com/pgsty/silo.pgsty.com) repository.

+## Contributor recognition
+
+Every human issue or pull-request author is recognized in [CONTRIBUTORS.md](CONTRIBUTORS.md),
+including open issues, draft PRs, and PRs closed without merging. Merged fixes,
+adopted proposals, and reports that lead to fixes receive greater prominence;
+first participation guides the remaining order. Work incorporated through a
+later PR retains credit without changing the original PR's recorded status.
+Security disclosures are credited with the reporter's agreement. DCO sign-off
+applies to code commits, not to opening an issue.
+
 ## Licensing of Contributions

 Code contributions to PGSTY SILO (`pgsty/silo`) are accepted under the
@@ -17,9 +17,9 @@ ENV GOPATH=/go
 ENV CGO_ENABLED=0

 ARG MC_REPO=pgsty/mc
-ARG MC_VERSION=RELEASE.2026-09-13T00-00-00Z
-ARG MC_AMD64_SHA256=9d2a92de9c7b887d9b944fe9ddce68d23f1b6df3415092e737594e56593f5e2b
-ARG MC_ARM64_SHA256=3d82e9ea6c601c4cb44fe5dd5f2ad1b7d7d64369378110f9ada9c524688a452a
+ARG MC_VERSION=RELEASE.2026-09-16T00-00-00Z
+ARG MC_AMD64_SHA256=4ba2814fd5507fbe6b4d237c359750b9119d28d7217495fa5b48002fcbd397ef
+ARG MC_ARM64_SHA256=b7008ca2a1bc5735b6585981c59a3640a0daa152dc789df3d1a0d0438787de82

 RUN apk add -U --no-cache \
    ca-certificates \
@@ -38,7 +38,7 @@
 ## Current release and main branch

 The latest published Server is [20260903](https://github.com/pgsty/silo/releases/tag/RELEASE.2026-09-03T13-18-01Z).
-As of 2026-09-13, the main branch has newer security, storage, Console and
+As of 2026-09-16, the main branch has newer security, storage, Console and
 shared-package changes that have not shipped in a Server release. See
 [CHANGELOG.md](CHANGELOG.md) and the [component version matrix](https://silo.pgsty.com/compatibility/versions/)
 for the exact release/source boundary, including SN-2026-011 and password-policy migration.
@@ -93,13 +93,16 @@ Every release ships checksums, SPDX SBOMs, Sigstore-signed manifests, and GitHub

 ## Compatibility

-The S3 API, `MINIO_*` variables, `minio_*` metrics, `x-minio-*` headers, `/minio/*` routes, the `github.com/minio/*` import paths, and the on-disk format (including `.minio.sys`) are preserved and held in place by a CI compatibility check. Only Silo-owned delivery surfaces change: the `silo` executable, package, service, Helm chart, and container image — no `minio` binary alias is installed.
+Silo preserves S3 and storage-format compatibility, including existing `MINIO_*` variables, `minio_*` metrics, `x-minio-*` headers, `/minio/*` routes, and `.minio.sys` data. CI guards selected compatibility identifiers; release notes document intentional security and behavior changes. Silo-owned delivery surfaces use the `silo` executable, package, service, Helm chart, and container image; no `minio` server binary alias is installed.
+
+The supported release stack is `pgsty/silo` + `pgsty/silo-console` + `pgsty/mc` + `pgsty/silo-pkg`; compatibility with unmodified upstream MinIO/MC is best effort. The Server, Console, and client retain their historical module paths where needed, while maintained code imports `github.com/pgsty/silo-pkg/v3` directly. The SDK `github.com/minio/minio-go/v7` is an explicit upstream dependency. See the current [go.mod](go.mod) for versions and replacements.

 Every divergence from upstream is listed in the code-verified [compatibility audit](https://silo.pgsty.com/compatibility/server/). Treat each release as a downstream upgrade: pin versions, read the [release notes](https://silo.pgsty.com/tags/silo/), and keep a rollback path.

 ### TLS and Go upgrades

-TLS key exchange follows Go's defaults across the S3 listener, node links,
+The following TLS repair is on main and is not included in Server 20260903.
+With that repair, TLS key exchange follows Go's defaults across the S3 listener, node links,
 replication, identity providers, etcd, and external HTTP services. If an endpoint
 cannot accept ML-KEM, `GODEBUG=tlsmlkem=0` disables the default hybrid exchanges
 for the process; certificate verification remains enabled. This option does not
@@ -115,7 +118,16 @@ values to restore Keychain trust. Explicit certificates in the configured `CAs`
 directory remain additive to the selected root pool.
 Go 1.27 binaries require macOS 13 or later. See the
 [Go release notes](https://go.dev/doc/go1.27) and the
-[SILO stack investigation](docs/investigations/go127-stack.md).
+[Go 1.27 TLS and OIDC discovery guide](https://silo.pgsty.com/blog/design/go127-tls-oidc-discovery/).
+
+## Documentation ownership
+
+User documentation is maintained at [silo.pgsty.com](https://silo.pgsty.com/docs/),
+with source in [pgsty/silo.pgsty.com](https://github.com/pgsty/silo.pgsty.com).
+The remaining `docs/` tree contains inherited references, examples, and tooling
+fixtures. Investigation logs, AI work records, and temporary reports are kept
+outside this repository; reusable findings belong in the companion site.
+See [AGENTS.md](AGENTS.md) for repository ownership and maintenance rules.

 ## Security & Contributing

@@ -123,53 +135,25 @@ Report vulnerabilities privately as described in [`SECURITY.md`](SECURITY.md); e

 ## Contributors

-**41 community contributors** build SILO, Console, mcli, shared packages, and related projects. The list includes maintainers and every human Issue or PR author, ordered by merged PRs, other PRs, then issue reports. Gold rings highlight significant contributions.
+<!-- Generated by silo.pgsty.com/bin/contributors.py. -->

-<p align="center">
-<a href="https://github.com/Vonng"><img src="https://silo.pgsty.com/images/contributors/Vonng.svg" width="60" height="60" alt="@Vonng" title="@Vonng — Maintains SILO, Console, mcli, shared packages, releases, and documentation"></a>
-<a href="https://github.com/h5vx"><img src="https://silo.pgsty.com/images/contributors/h5vx.svg" width="60" height="60" alt="@h5vx" title="@h5vx — Implemented per-bucket CORS configuration and enforcement"></a>
-<a href="https://github.com/mrjavadseydi"><img src="https://silo.pgsty.com/images/contributors/mrjavadseydi.svg" width="60" height="60" alt="@mrjavadseydi" title="@mrjavadseydi — Fixed effective bucket quota metrics; proposed access-frequency ILM"></a>
-<a href="https://github.com/Dansyuqri"><img src="https://silo.pgsty.com/images/contributors/Dansyuqri.svg" width="60" height="60" alt="@Dansyuqri" title="@Dansyuqri — Added ChecksumType to multipart completion responses"></a>
-<a href="https://github.com/ycjlin"><img src="https://silo.pgsty.com/images/contributors/ycjlin.svg" width="60" height="60" alt="@ycjlin" title="@ycjlin — Fixed missing-bucket ListObjects semantics"></a>
-<a href="https://github.com/pinginfo"><img src="https://silo.pgsty.com/images/contributors/pinginfo.svg" width="60" height="60" alt="@pinginfo" title="@pinginfo — Repaired bucket notification streaming"></a>
-<a href="https://github.com/ZouhairCharef"><img src="https://silo.pgsty.com/images/contributors/ZouhairCharef.svg" width="60" height="60" alt="@ZouhairCharef" title="@ZouhairCharef — Patched CVE-2026-34986 in go-jose"></a>
-<a href="https://github.com/mfredenhagen"><img src="https://silo.pgsty.com/images/contributors/mfredenhagen.svg" width="60" height="60" alt="@mfredenhagen" title="@mfredenhagen — Patched CVE-2026-39883 in OpenTelemetry"></a>
-<a href="https://github.com/waterkip"><img src="https://silo.pgsty.com/images/contributors/waterkip.svg" width="60" height="60" alt="@waterkip" title="@waterkip — Repointed documentation links to the SILO portal"></a>
-<a href="https://github.com/mikemikimike"><img src="https://silo.pgsty.com/images/contributors/mikemikimike.svg" width="60" height="60" alt="@mikemikimike" title="@mikemikimike — Contributed the replicated SSE-C plaintext part-size fix"></a>
-<a href="https://github.com/metaneutrons"><img src="https://silo.pgsty.com/images/contributors/metaneutrons.svg" width="60" height="60" alt="@metaneutrons" title="@metaneutrons — Reported and proposed explicit-version delete authorization"></a>
-<a href="https://github.com/magicxor"><img src="https://silo.pgsty.com/images/contributors/magicxor.svg" width="60" height="60" alt="@magicxor" title="@magicxor — Reported and proposed conditional DELETE support for If-Match"></a>
-<a href="https://github.com/davinkevin"><img src="https://silo.pgsty.com/images/contributors/davinkevin.svg" width="60" height="60" alt="@davinkevin" title="@davinkevin — Proposed the distroless container image and dependency automation"></a>
-<a href="https://github.com/lem21h"><img src="https://silo.pgsty.com/images/contributors/lem21h.svg" width="48" height="48" alt="@lem21h" title="@lem21h — Proposed robustness and goroutine improvements"></a>
-<a href="https://github.com/sulin37392"><img src="https://silo.pgsty.com/images/contributors/sulin37392.svg" width="48" height="48" alt="@sulin37392" title="@sulin37392 — Proposed dependency updates"></a>
-<a href="https://github.com/cbornet"><img src="https://silo.pgsty.com/images/contributors/cbornet.svg" width="60" height="60" alt="@cbornet" title="@cbornet — Reported multipart and streaming checksum defects and missing-bucket semantics"></a>
-<a href="https://github.com/vampywiz17"><img src="https://silo.pgsty.com/images/contributors/vampywiz17.svg" width="60" height="60" alt="@vampywiz17" title="@vampywiz17 — Reported LDAP TLS and Console login regressions"></a>
-<a href="https://github.com/orenyomtov"><img src="https://silo.pgsty.com/images/contributors/orenyomtov.svg" width="60" height="60" alt="@orenyomtov" title="@orenyomtov — Reported the unsigned-header CopyObject cross-object read (SN-2026-011)"></a>
-<a href="https://github.com/mumu-lab"><img src="https://silo.pgsty.com/images/contributors/mumu-lab.svg" width="48" height="48" alt="@mumu-lab" title="@mumu-lab — Reported bucket quota metrics reading a deprecated field"></a>
-<a href="https://github.com/jvasile"><img src="https://silo.pgsty.com/images/contributors/jvasile.svg" width="48" height="48" alt="@jvasile" title="@jvasile — Reported missing user, group, and defaults in Debian packages"></a>
-<a href="https://github.com/pmezhuev"><img src="https://silo.pgsty.com/images/contributors/pmezhuev.svg" width="48" height="48" alt="@pmezhuev" title="@pmezhuev — Reported missing RPM package signatures"></a>
-<a href="https://github.com/TLINDEN"><img src="https://silo.pgsty.com/images/contributors/TLINDEN.svg" width="48" height="48" alt="@TLINDEN" title="@TLINDEN — Reported the missing client in release tarballs"></a>
-<a href="https://github.com/makinikm"><img src="https://silo.pgsty.com/images/contributors/makinikm.svg" width="48" height="48" alt="@makinikm" title="@makinikm — Reported the missing client in the container image"></a>
-<a href="https://github.com/meesudzu"><img src="https://silo.pgsty.com/images/contributors/meesudzu.svg" width="48" height="48" alt="@meesudzu" title="@meesudzu — Requested the migration guide from upstream MinIO"></a>
-<a href="https://github.com/kuldeep-link11"><img src="https://silo.pgsty.com/images/contributors/kuldeep-link11.svg" width="48" height="48" alt="@kuldeep-link11" title="@kuldeep-link11 — Reported NATS JWT credentials and target reload issues"></a>
-<a href="https://github.com/sargarass"><img src="https://silo.pgsty.com/images/contributors/sargarass.svg" width="48" height="48" alt="@sargarass" title="@sargarass — Reported ListMultipartUploads prefix and pagination semantics"></a>
-<a href="https://github.com/liuhaodongliu990-cmyk"><img src="https://silo.pgsty.com/images/contributors/liuhaodongliu990-cmyk.svg" width="48" height="48" alt="@liuhaodongliu990-cmyk" title="@liuhaodongliu990-cmyk — Reported indeterminate progress for prefix downloads"></a>
-<a href="https://github.com/Xavier-777"><img src="https://silo.pgsty.com/images/contributors/Xavier-777.svg" width="48" height="48" alt="@Xavier-777" title="@Xavier-777 — Reported Console lifecycle management and file preview gaps"></a>
-<a href="https://github.com/spaceg00se-r"><img src="https://silo.pgsty.com/images/contributors/spaceg00se-r.svg" width="48" height="48" alt="@spaceg00se-r" title="@spaceg00se-r — Requested cpuv1 support and reported a workflow token failure"></a>
-<a href="https://github.com/kh0mka"><img src="https://silo.pgsty.com/images/contributors/kh0mka.svg" width="48" height="48" alt="@kh0mka" title="@kh0mka — Reported inter-node I/O timeouts in ReadFileStreamHandler"></a>
-<a href="https://github.com/bagutzu"><img src="https://silo.pgsty.com/images/contributors/bagutzu.svg" width="48" height="48" alt="@bagutzu" title="@bagutzu — Requested KES-compatible external KMS and OpenBao support"></a>
-<a href="https://github.com/DestroyLee"><img src="https://silo.pgsty.com/images/contributors/DestroyLee.svg" width="48" height="48" alt="@DestroyLee" title="@DestroyLee — Reported the missing documentation navigation"></a>
-<a href="https://github.com/mosesdd"><img src="https://silo.pgsty.com/images/contributors/mosesdd.svg" width="48" height="48" alt="@mosesdd" title="@mosesdd — Requested a maintained Helm chart"></a>
-<a href="https://github.com/zylpsrs"><img src="https://silo.pgsty.com/images/contributors/zylpsrs.svg" width="48" height="48" alt="@zylpsrs" title="@zylpsrs — Reported missing Console tiering and site replication"></a>
-<a href="https://github.com/heroes1412"><img src="https://silo.pgsty.com/images/contributors/heroes1412.svg" width="48" height="48" alt="@heroes1412" title="@heroes1412 — Reported the unusable profiling option"></a>
-<a href="https://github.com/redfoxfox"><img src="https://silo.pgsty.com/images/contributors/redfoxfox.svg" width="48" height="48" alt="@redfoxfox" title="@redfoxfox — Reported Chinese documentation availability"></a>
-<a href="https://github.com/jiadzh"><img src="https://silo.pgsty.com/images/contributors/jiadzh.svg" width="48" height="48" alt="@jiadzh" title="@jiadzh — Requested Windows build guidance"></a>
-<a href="https://github.com/AntonOfTheWoods"><img src="https://silo.pgsty.com/images/contributors/AntonOfTheWoods.svg" width="48" height="48" alt="@AntonOfTheWoods" title="@AntonOfTheWoods — Asked for clarity on Helm chart and operator options"></a>
-<a href="https://github.com/chalukyaj"><img src="https://silo.pgsty.com/images/contributors/chalukyaj.svg" width="48" height="48" alt="@chalukyaj" title="@chalukyaj — Proposed making the SILO Operator easier to discover"></a>
-<a href="https://github.com/nsanitate"><img src="https://silo.pgsty.com/images/contributors/nsanitate.svg" width="48" height="48" alt="@nsanitate" title="@nsanitate — Proposed CNCF Sandbox governance"></a>
-<a href="https://github.com/Kesavaambati"><img src="https://silo.pgsty.com/images/contributors/Kesavaambati.svg" width="48" height="48" alt="@Kesavaambati" title="@Kesavaambati — Asked about community support and image maintenance"></a>
-</p>
+Every human issue or pull-request author is part of the SILO community, including open and unmerged work. Merged fixes, adopted proposals, and actionable reports receive priority, with first participation guiding the remaining order. Gold rings highlight reviewed significant contributions.

-[View the full contribution record](CONTRIBUTORS.md) for each person's proposals, fixes, and reports.
+<a href="CONTRIBUTORS.md">
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/contributors-dark.svg">
+  <img src="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/contributors-light.svg" alt="SILO community contributors">
+</picture>
+</a>
+
+[View contribution notes and actual PR status](CONTRIBUTORS.md).
+
+## Star History
+
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/star-history-dark.svg">
+  <img src="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/star-history-light.svg" alt="SILO GitHub star history">
+</picture>

 ## Background

@@ -38,7 +38,7 @@
 ## 当前发行版与主分支

 最新已发布的 Server 仍为 [20260903](https://github.com/pgsty/silo/releases/tag/RELEASE.2026-09-03T13-18-01Z)。
-截至 2026-09-13，主分支已合入更新的安全、存储、Console 与共享包改动，但尚未发布新 Server。
+截至 2026-09-16，主分支已合入更新的安全、存储、Console 与共享包改动，但尚未发布新 Server。
 准确的已发布/源码边界见 [CHANGELOG.md](CHANGELOG.md) 与[组件版本矩阵](https://silo.pgsty.com/zh/compatibility/versions/)，
 其中包括 SN-2026-011 修复状态与密码权限迁移要求。

@@ -92,63 +92,50 @@ docker exec silo mcli mb local/demo && docker exec silo mcli ls local

 ## 兼容性

-S3 API、`MINIO_*` 环境变量、`minio_*` 指标、`x-minio-*` 头、`/minio/*` 路由、`github.com/minio/*` 导入路径与磁盘格式（含 `.minio.sys`）原样保留，并由 CI 兼容性门禁冻结。只有 Silo 自有交付面改名：`silo` 可执行文件、软件包、服务、Helm Chart 与容器镜像 —— 原生交付物不会安装 `minio` 二进制别名。
+Silo 保留 S3 与存储格式兼容性，包括既有 `MINIO_*` 环境变量、`minio_*` 指标、`x-minio-*` 头、`/minio/*` 路由与 `.minio.sys` 数据。CI 守卫检查选定的兼容性标识，有意的安全与行为变化在发布说明中记录。Silo 自有交付面使用 `silo` 可执行文件、软件包、服务、Helm Chart 与容器镜像；原生交付物不会安装 `minio` 服务端二进制别名。
+
+正式支持和发布验收的组合为 `pgsty/silo` + `pgsty/silo-console` + `pgsty/mc` + `pgsty/silo-pkg`，对未修改的上游 MinIO/MC 尽最大努力保持兼容。Server、Console 与客户端按需保留历史模块路径，维护源码直接导入 `github.com/pgsty/silo-pkg/v3`；SDK `github.com/minio/minio-go/v7` 是明确保留的上游依赖。具体版本与 replace 以当前 [go.mod](go.mod) 为准。

 与上游的全部分歧，以逐项核验代码的[兼容性审计](https://silo.pgsty.com/zh/compatibility/server/)形式维护。每个版本仍应视为下游升级：锁定版本，阅读[版本说明](https://silo.pgsty.com/zh/tags/silo/)，并保留回滚路径。

+### TLS 与 Go 升级
+
+Server 恢复 Go 默认密钥交换策略的修复已在 main，尚未包含在 Server 20260903。
+`GODEBUG=tlsmlkem=0`、`tlssecpmlkem=0` 的适用范围、macOS 根证书来源变化，
+以及 OIDC discovery 的诊断方法见 [Go 1.27 TLS 与 OIDC 指南](https://silo.pgsty.com/zh/blog/design/go127-tls-oidc-discovery/)。
+
+## 文档归属
+
+用户文档统一维护在 [silo.pgsty.com](https://silo.pgsty.com/zh/docs/)，源码位于
+[pgsty/silo.pgsty.com](https://github.com/pgsty/silo.pgsty.com)。本仓库保留的 `docs/`
+主要是继承的参考材料、示例和工具测试夹具。调查日志、AI 工作记录与临时报告放在仓库外；
+可复用的结论应整理进伴生文档站。仓库职责与维护规则见 [AGENTS.md](AGENTS.md)。
+
 ## 安全与贡献

 请按照 [`SECURITY.md`](SECURITY.md) 私密报告漏洞；每项修复都会发布公开[安全公告](https://silo.pgsty.com/zh/blog/security/)。本项目不要求签署 CLA：贡献按 AGPL-3.0-or-later（inbound=outbound）接收，只需 DCO 签署（`git commit -s`），详见 [`CONTRIBUTING.md`](CONTRIBUTING.md)。

 ## 贡献者

-**41 位社区贡献者**共同建设 SILO、Console、mcli、公共包与相关项目。名单包含维护者，以及所有提出 Issue 或 PR 的真人作者；按已合并 PR、其他 PR、Issue 报告排序，黄圈标记显著贡献。
+<!-- Generated by silo.pgsty.com/bin/contributors.py. -->

-<p align="center">
-<a href="https://github.com/Vonng"><img src="https://silo.pgsty.com/images/contributors/Vonng.svg" width="60" height="60" alt="@Vonng" title="@Vonng — 维护 SILO、Console、mcli、公共包、发行与文档"></a>
-<a href="https://github.com/h5vx"><img src="https://silo.pgsty.com/images/contributors/h5vx.svg" width="60" height="60" alt="@h5vx" title="@h5vx — 实现单桶 CORS 配置与请求执行"></a>
-<a href="https://github.com/mrjavadseydi"><img src="https://silo.pgsty.com/images/contributors/mrjavadseydi.svg" width="60" height="60" alt="@mrjavadseydi" title="@mrjavadseydi — 修复有效桶配额指标，并提交按访问频率分层的 ILM 方案"></a>
-<a href="https://github.com/Dansyuqri"><img src="https://silo.pgsty.com/images/contributors/Dansyuqri.svg" width="60" height="60" alt="@Dansyuqri" title="@Dansyuqri — 为分片上传完成响应补充 ChecksumType"></a>
-<a href="https://github.com/ycjlin"><img src="https://silo.pgsty.com/images/contributors/ycjlin.svg" width="60" height="60" alt="@ycjlin" title="@ycjlin — 修复缺失桶的 ListObjects 语义"></a>
-<a href="https://github.com/pinginfo"><img src="https://silo.pgsty.com/images/contributors/pinginfo.svg" width="60" height="60" alt="@pinginfo" title="@pinginfo — 修复桶通知的流式输出"></a>
-<a href="https://github.com/ZouhairCharef"><img src="https://silo.pgsty.com/images/contributors/ZouhairCharef.svg" width="60" height="60" alt="@ZouhairCharef" title="@ZouhairCharef — 修复 go-jose 中的 CVE-2026-34986"></a>
-<a href="https://github.com/mfredenhagen"><img src="https://silo.pgsty.com/images/contributors/mfredenhagen.svg" width="60" height="60" alt="@mfredenhagen" title="@mfredenhagen — 修复 OpenTelemetry 中的 CVE-2026-39883"></a>
-<a href="https://github.com/waterkip"><img src="https://silo.pgsty.com/images/contributors/waterkip.svg" width="60" height="60" alt="@waterkip" title="@waterkip — 将文档链接指向 SILO 门户"></a>
-<a href="https://github.com/mikemikimike"><img src="https://silo.pgsty.com/images/contributors/mikemikimike.svg" width="60" height="60" alt="@mikemikimike" title="@mikemikimike — 提交 SSE-C 复制分片明文尺寸修复"></a>
-<a href="https://github.com/metaneutrons"><img src="https://silo.pgsty.com/images/contributors/metaneutrons.svg" width="60" height="60" alt="@metaneutrons" title="@metaneutrons — 报告并提交显式版本删除鉴权方案"></a>
-<a href="https://github.com/magicxor"><img src="https://silo.pgsty.com/images/contributors/magicxor.svg" width="60" height="60" alt="@magicxor" title="@magicxor — 报告并提交 DELETE If-Match 条件请求支持方案"></a>
-<a href="https://github.com/davinkevin"><img src="https://silo.pgsty.com/images/contributors/davinkevin.svg" width="60" height="60" alt="@davinkevin" title="@davinkevin — 提交 distroless 容器镜像与依赖自动更新方案"></a>
-<a href="https://github.com/lem21h"><img src="https://silo.pgsty.com/images/contributors/lem21h.svg" width="48" height="48" alt="@lem21h" title="@lem21h — 提交健壮性与 goroutine 改进"></a>
-<a href="https://github.com/sulin37392"><img src="https://silo.pgsty.com/images/contributors/sulin37392.svg" width="48" height="48" alt="@sulin37392" title="@sulin37392 — 提交依赖更新"></a>
-<a href="https://github.com/cbornet"><img src="https://silo.pgsty.com/images/contributors/cbornet.svg" width="60" height="60" alt="@cbornet" title="@cbornet — 报告分片与流式校验和缺陷及缺失桶语义问题"></a>
-<a href="https://github.com/vampywiz17"><img src="https://silo.pgsty.com/images/contributors/vampywiz17.svg" width="60" height="60" alt="@vampywiz17" title="@vampywiz17 — 报告 LDAP TLS 与 Console 登录回归"></a>
-<a href="https://github.com/orenyomtov"><img src="https://silo.pgsty.com/images/contributors/orenyomtov.svg" width="60" height="60" alt="@orenyomtov" title="@orenyomtov — 报告未签名头导致的 CopyObject 跨对象读取（SN-2026-011）"></a>
-<a href="https://github.com/mumu-lab"><img src="https://silo.pgsty.com/images/contributors/mumu-lab.svg" width="48" height="48" alt="@mumu-lab" title="@mumu-lab — 报告桶配额指标读取已弃用字段的问题"></a>
-<a href="https://github.com/jvasile"><img src="https://silo.pgsty.com/images/contributors/jvasile.svg" width="48" height="48" alt="@jvasile" title="@jvasile — 报告 Debian 包缺少用户、用户组与默认配置"></a>
-<a href="https://github.com/pmezhuev"><img src="https://silo.pgsty.com/images/contributors/pmezhuev.svg" width="48" height="48" alt="@pmezhuev" title="@pmezhuev — 报告 RPM 包缺少 GPG 签名"></a>
-<a href="https://github.com/TLINDEN"><img src="https://silo.pgsty.com/images/contributors/TLINDEN.svg" width="48" height="48" alt="@TLINDEN" title="@TLINDEN — 报告发布压缩包缺少客户端"></a>
-<a href="https://github.com/makinikm"><img src="https://silo.pgsty.com/images/contributors/makinikm.svg" width="48" height="48" alt="@makinikm" title="@makinikm — 报告容器镜像缺少客户端"></a>
-<a href="https://github.com/meesudzu"><img src="https://silo.pgsty.com/images/contributors/meesudzu.svg" width="48" height="48" alt="@meesudzu" title="@meesudzu — 提出从上游 MinIO 迁移的指南需求"></a>
-<a href="https://github.com/kuldeep-link11"><img src="https://silo.pgsty.com/images/contributors/kuldeep-link11.svg" width="48" height="48" alt="@kuldeep-link11" title="@kuldeep-link11 — 报告 NATS JWT 凭据与通知目标重载问题"></a>
-<a href="https://github.com/sargarass"><img src="https://silo.pgsty.com/images/contributors/sargarass.svg" width="48" height="48" alt="@sargarass" title="@sargarass — 报告 ListMultipartUploads 前缀与分页语义问题"></a>
-<a href="https://github.com/liuhaodongliu990-cmyk"><img src="https://silo.pgsty.com/images/contributors/liuhaodongliu990-cmyk.svg" width="48" height="48" alt="@liuhaodongliu990-cmyk" title="@liuhaodongliu990-cmyk — 报告前缀下载进度显示异常"></a>
-<a href="https://github.com/Xavier-777"><img src="https://silo.pgsty.com/images/contributors/Xavier-777.svg" width="48" height="48" alt="@Xavier-777" title="@Xavier-777 — 报告 Console 生命周期管理与文件预览缺失"></a>
-<a href="https://github.com/spaceg00se-r"><img src="https://silo.pgsty.com/images/contributors/spaceg00se-r.svg" width="48" height="48" alt="@spaceg00se-r" title="@spaceg00se-r — 提出 cpuv1 支持需求并报告工作流令牌错误"></a>
-<a href="https://github.com/kh0mka"><img src="https://silo.pgsty.com/images/contributors/kh0mka.svg" width="48" height="48" alt="@kh0mka" title="@kh0mka — 报告 ReadFileStreamHandler 节点间 I/O 超时"></a>
-<a href="https://github.com/bagutzu"><img src="https://silo.pgsty.com/images/contributors/bagutzu.svg" width="48" height="48" alt="@bagutzu" title="@bagutzu — 提出兼容 KES 的外部 KMS 与 OpenBao 支持需求"></a>
-<a href="https://github.com/DestroyLee"><img src="https://silo.pgsty.com/images/contributors/DestroyLee.svg" width="48" height="48" alt="@DestroyLee" title="@DestroyLee — 报告文档目录导航缺失"></a>
-<a href="https://github.com/mosesdd"><img src="https://silo.pgsty.com/images/contributors/mosesdd.svg" width="48" height="48" alt="@mosesdd" title="@mosesdd — 提出维护 Helm Chart 的需求"></a>
-<a href="https://github.com/zylpsrs"><img src="https://silo.pgsty.com/images/contributors/zylpsrs.svg" width="48" height="48" alt="@zylpsrs" title="@zylpsrs — 报告 Console 缺少分层与站点复制"></a>
-<a href="https://github.com/heroes1412"><img src="https://silo.pgsty.com/images/contributors/heroes1412.svg" width="48" height="48" alt="@heroes1412" title="@heroes1412 — 报告性能分析选项不可用"></a>
-<a href="https://github.com/redfoxfox"><img src="https://silo.pgsty.com/images/contributors/redfoxfox.svg" width="48" height="48" alt="@redfoxfox" title="@redfoxfox — 报告中文文档站点不可用"></a>
-<a href="https://github.com/jiadzh"><img src="https://silo.pgsty.com/images/contributors/jiadzh.svg" width="48" height="48" alt="@jiadzh" title="@jiadzh — 提出 Windows 构建指导需求"></a>
-<a href="https://github.com/AntonOfTheWoods"><img src="https://silo.pgsty.com/images/contributors/AntonOfTheWoods.svg" width="48" height="48" alt="@AntonOfTheWoods" title="@AntonOfTheWoods — 提出明确 Helm Chart 与 Operator 选项的需求"></a>
-<a href="https://github.com/chalukyaj"><img src="https://silo.pgsty.com/images/contributors/chalukyaj.svg" width="48" height="48" alt="@chalukyaj" title="@chalukyaj — 提出改善 SILO Operator 可发现性的建议"></a>
-<a href="https://github.com/nsanitate"><img src="https://silo.pgsty.com/images/contributors/nsanitate.svg" width="48" height="48" alt="@nsanitate" title="@nsanitate — 提出加入 CNCF Sandbox 的治理建议"></a>
-<a href="https://github.com/Kesavaambati"><img src="https://silo.pgsty.com/images/contributors/Kesavaambati.svg" width="48" height="48" alt="@Kesavaambati" title="@Kesavaambati — 提出社区支持与容器镜像维护问题"></a>
-</p>
+每位 issue 或 PR 的作者都是 SILO 社区的一员，包括尚未合并的工作。已合并的修复、被采纳的方案和有效报告优先展示，其余参考首次参与时间；金色圆环突出经过审核的显著贡献。

-[查看完整贡献记录](CONTRIBUTORS.md)，了解每位贡献者的提案、修复与问题报告。
+<a href="CONTRIBUTORS.md">
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/contributors-dark.svg">
+  <img src="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/contributors-light.svg" alt="SILO 社区贡献者">
+</picture>
+</a>
+
+[查看贡献记录与实际 PR 状态](CONTRIBUTORS.md)。
+
+## Star History
+
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/star-history-dark.svg">
+  <img src="https://raw.githubusercontent.com/pgsty/silo/codex/repository-cards/star-history-light.svg" alt="SILO GitHub 星标历史">
+</picture>

 ## 背景

@@ -7,7 +7,7 @@ Silo-specific fixes or release notes.
 ## Supported Versions

 Security fixes are tracked on the active development branch and summarized in
-[docs/security/advisories.md](docs/security/advisories.md). Only the current
+[the security advisory ledger](https://silo.pgsty.com/about/security-advisories/). Only the current
 Silo release line is supported unless an advisory says otherwise.

 ## Inherited Fix Evidence
@@ -26,7 +26,7 @@ separately even when the fork preserves the original commit object and SHA.
  The inherited [service-account](https://github.com/pgsty/silo/blob/c1a49490c78e9c3ebcad86ba0662319138ace190/cmd/admin-handlers-users_test.go#L211-L212)
  and [STS](https://github.com/pgsty/silo/blob/c1a49490c78e9c3ebcad86ba0662319138ace190/cmd/sts-handlers_test.go#L45-L46)
  regression groups remain part of `go test ./cmd`; see the
-  [canonical ledger](docs/security/advisories.md#inherited-upstream-advisory-baseline)
+  [canonical ledger](https://silo.pgsty.com/about/security-advisories/#inherited)
  for the operator-facing record.

 ## Reporting a Vulnerability
@@ -42,4 +42,4 @@ For vulnerabilities in this fork:

 ## Disclosure Process

-Fork-specific fixes and user-visible upgrade notes are published in [docs/security/advisories.md](docs/security/advisories.md). The fork-specific triage and remediation process is described in [VULNERABILITY_REPORT.md](VULNERABILITY_REPORT.md).
+Fork-specific fixes and user-visible upgrade notes are published in [the security advisory ledger](https://silo.pgsty.com/about/security-advisories/). The fork-specific triage and remediation process is described in [VULNERABILITY_REPORT.md](VULNERABILITY_REPORT.md).
@@ -34,4 +34,4 @@ Based on the report, the Silo maintainers investigate:
 If the vulnerability exists in this fork itself, the maintainers will, when
 feasible, fix the issue or implement reasonable countermeasures such that the
 vulnerability can no longer be exploited. Fork-specific upgrade notes and
-security advisories are published in `docs/security/advisories.md`.
+security advisories are published in [the security advisory ledger](https://silo.pgsty.com/about/security-advisories/).
@@ -40,7 +40,7 @@ if [ -n "${MCLI_BIN:-}" ]; then
 	exit 0
 fi

-release=${MCLI_RELEASE:-RELEASE.2026-09-13T00-00-00Z}
+release=${MCLI_RELEASE:-RELEASE.2026-09-16T00-00-00Z}
 version_hyphen=${release#RELEASE.}
 package_version=$(printf '%s\n' "${version_hyphen}" | sed -E 's/^([0-9]{4})-([0-9]{2})-([0-9]{2})T([0-9]{2})-([0-9]{2})-([0-9]{2})Z$/\1\2\3\4\5\6.0.0/')
 if [ "${package_version}" = "${version_hyphen}" ]; then
@@ -132,8 +132,8 @@
    "MINIO_API_DELETE_CLEANUP_INTERVAL",
    "MINIO_API_DISABLE_ODIRECT",
    "MINIO_API_GZIP_OBJECTS",
-    "MINIO_API_LEGACY_BUCKET_RESOURCE_MATCH",
    "MINIO_API_LIST_QUORUM",
+    "MINIO_API_MULTIPART_LISTING",
    "MINIO_API_OBJECT_MAX_VERSIONS",
    "MINIO_API_ODIRECT",
    "MINIO_API_REMOTE_TRANSPORT_DEADLINE",
@@ -571,6 +571,14 @@
    "minio_resource_stats",
    "minio_s3",
    "minio_stats",
+    "minio_system_cpu_avg_idle",
+    "minio_system_cpu_avg_iowait",
+    "minio_system_cpu_load",
+    "minio_system_cpu_load_perc",
+    "minio_system_cpu_nice",
+    "minio_system_cpu_steal",
+    "minio_system_cpu_system",
+    "minio_system_cpu_user",
    "minio_test"
  ],
  "headers": [
@@ -627,7 +635,6 @@
    "x-minio-origin-endpoint",
    "x-minio-prefixes-total",
    "x-minio-read-quorum",
-    "x-minio-replication",
    "x-minio-replication-actual-object-size",
    "x-minio-replication-delete-status",
    "x-minio-replication-deletemarker-status",
@@ -766,6 +773,7 @@
    "/metrics/v3",
    "/minio/grid/",
    "/minio/grid/lock/",
+    "/multipart-preflight",
    "/netperf",
    "/notification",
    "/oauth2/callback",
@@ -115,9 +115,8 @@ func collect(repo string) (manifest, error) {
 	fset := token.NewFileSet()

 	for _, rel := range files {
-		// Investigation artifacts contain synthetic routes and archived configurations.
+		// Migration notes and guard fixtures contain archived identifiers.
 		if rel == "SILO_REBRANDING_MIGRATION.md" ||
-			strings.HasPrefix(rel, "docs/investigations/") ||
 			strings.HasPrefix(rel, "buildscripts/rebrand-guard/") ||
 			strings.HasPrefix(rel, "buildscripts/helm-migration-guard/") {
 			continue
@@ -0,0 +1 @@
+__pycache__/
@@ -0,0 +1,163 @@
+# Copyright (c) 2026 Feng Ruohang
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+"""Pure, self-contained SVG rendering for SILO's README cards."""
+from datetime import date, timedelta
+from html import escape
+import math
+import xml.etree.ElementTree as ET
+
+themes = {
+    'light': dict(bg='#ffffff', wash='#f2f7fc', edge='#d9e3ee', ink='#16222e',
+                  muted='#62758a', blue='#1d588c', copper='#b4762e', grid='#e5edf5',
+                  line='#2b6ca3', ring='#dce5ef', field='#f7f9fc', label='#3d4e61'),
+    'dark': dict(bg='#101923', wash='#152738', edge='#2b3c50', ink='#e8eef6',
+                 muted='#93a3b8', blue='#7fb8e8', copper='#e0a35c', grid='#263749',
+                 line='#5da2dd', ring='#3a4e63', field='#0b1119', label='#b6c2d2'),
+}
+
+def read_emblem(path):
+    emblem = ET.parse(path).getroot()
+    body = ''.join(ET.tostring(child, encoding='unicode') for child in emblem
+                   if child.tag.rsplit('}', 1)[-1] in ('defs', 'g'))
+    return '\n'.join(line.rstrip() for line in body.splitlines()).strip()
+
+
+def txt(x, y, value, size=14, color=None, weight=400, anchor='start', mono=False, spacing=None):
+    family = 'Menlo,Consolas,monospace' if mono else 'Arial,Helvetica,sans-serif'
+    extra = f' letter-spacing="{spacing}"' if spacing is not None else ''
+    return (f'<text x="{x}" y="{y}" font-family="{family}" font-size="{size}" '
+            f'font-weight="{weight}" fill="{color}" text-anchor="{anchor}"{extra}>'
+            f'{escape(str(value))}</text>')
+
+
+def start(height, theme, title, description, emblem_body):
+    t = themes[theme]
+    return [f'<svg xmlns="http://www.w3.org/2000/svg" width="1000" height="{height}" '
+            f'viewBox="0 0 1000 {height}" role="img" aria-labelledby="title desc">',
+            f'<title id="title">{escape(title)}</title><desc id="desc">{escape(description)}</desc>',
+            '<defs><linearGradient id="surface" x1="0" y1="1" x2="1" y2="0">'
+            f'<stop offset="0" stop-color="{t["bg"]}"/>'
+            f'<stop offset="1" stop-color="{t["wash"]}"/></linearGradient>'
+            '<linearGradient id="accent" x1="0" y1="0" x2="1" y2="0">'
+            f'<stop offset="0" stop-color="{t["blue"]}"/>'
+            f'<stop offset="1" stop-color="{t["copper"]}"/></linearGradient>'
+            '<linearGradient id="area" x1="0" y1="0" x2="0" y2="1">'
+            f'<stop offset="0" stop-color="{t["line"]}" stop-opacity=".22"/>'
+            f'<stop offset="1" stop-color="{t["line"]}" stop-opacity=".015"/>'
+            '</linearGradient></defs>',
+            f'<rect x=".75" y=".75" width="998.5" height="{height-1.5}" rx="22" '
+            f'fill="url(#surface)" stroke="{t["edge"]}" stroke-width="1.5"/>',
+            f'<svg x="40" y="25" width="25" height="25" viewBox="230 213 570 570">{emblem_body}</svg>']
+
+
+def heading(parts, t, eyebrow, title, subtitle, value, value_label):
+    parts.extend([
+        txt(76, 43, eyebrow, 11, t['muted'], 600, mono=True, spacing=1.7),
+        txt(40, 94, title, 32, t['ink'], 700),
+        txt(41, 123, subtitle, 14, t['muted']),
+        txt(958, 89, f'{value:,}', 45, t['ink'], 700, anchor='end'),
+        txt(957, 114, value_label, 10, t['muted'], 600, anchor='end', mono=True, spacing=1.5),
+        f'<path d="M40 146 H960" stroke="{t["edge"]}"/>',
+    ])
+
+
+def contributors(theme, people, snapshot, emblem_body):
+    height = 206 + 76 * math.ceil(len(people) / 10)
+    t = themes[theme]
+    parts = start(height, theme, f'SILO community — {len(people)} contributors',
+                  f'The existing SILO community roll, including code, proposals and reports across related projects. '
+                  f'Gold rings retain the existing significant-contribution designation. Snapshot {snapshot}.', emblem_body)
+    heading(parts, t, 'SILO / COMMUNITY', 'Contributors',
+            'Code, proposals & reports across SILO and related projects', len(people), 'COMMUNITY CONTRIBUTORS')
+    for row in range(math.ceil(len(people) / 10)):
+        group = people[row * 10:(row + 1) * 10]
+        row_width = len(group) * 91
+        for col, person in enumerate(group):
+            x = (1000 - row_width) / 2 + col * 91 + 45.5
+            y = 199 + row * 76
+            identifier = f'avatar-{row}-{col}'
+            featured = bool(person.get('featured'))
+            parts.append(f'<g><title>@{escape(person["handle"])} — {escape(person["what"])}</title>')
+            parts.append(f'<defs><clipPath id="{identifier}"><circle cx="{x}" cy="{y}" r="29"/></clipPath></defs>')
+            if featured:
+                parts.append(f'<circle cx="{x}" cy="{y}" r="34" fill="{t["copper"]}" opacity=".09"/>')
+            if person.get('avatarDataUrl'):
+                parts.append(f'<image x="{x-29}" y="{y-29}" width="58" height="58" '
+                             f'clip-path="url(#{identifier})" href="{escape(person["avatarDataUrl"])}"/>')
+            else:
+                parts.append(f'<circle cx="{x}" cy="{y}" r="29" fill="{t["ring"]}"/>')
+                parts.append(txt(x, y + 9, person['handle'][0].upper(), 26, t['ink'], 700, 'middle'))
+            parts.append(f'<circle cx="{x}" cy="{y}" r="30.5" fill="none" '
+                         f'stroke="{t["copper"] if featured else t["ring"]}" stroke-width="{2 if featured else 1.25}"/></g>')
+    parts.extend([
+        f'<path d="M40 {height-42} H960" stroke="{t["edge"]}"/>',
+        f'<circle cx="47" cy="{height-21}" r="4" fill="none" stroke="{t["copper"]}" stroke-width="1.5"/>',
+        txt(61, height-17, 'Gold rings mark significant contributions', 12, t['muted']),
+        txt(959, height-17, f'AS OF {snapshot}', 10, t['muted'], 500, 'end', mono=True, spacing=.6),
+        '</svg>',
+    ])
+    return ''.join(parts)
+
+
+def stars(theme, history, snapshot, emblem_body):
+    points = history['points']
+    star_count = points[-1]['stars']
+    t = themes[theme]
+    provenance = ('Initial history reconstructed · Daily totals since ' + history['bootstrap']['through'] + ' · UTC'
+                  if history['bootstrap']['reconstructed'] else 'Observed daily star totals · UTC')
+    parts = start(558, theme, f'SILO star history — {star_count:,} stars',
+                  f'GitHub repository pgsty/silo. {star_count:,} stars as of {snapshot}. ' +
+                  provenance, emblem_body)
+    heading(parts, t, 'SILO / GITHUB', 'Star History', 'pgsty/silo', star_count, 'GITHUB STARS')
+    left, right, top, bottom = 76, 958, 177, 440
+    begin = date.fromisoformat(points[0]['date'])
+    end = date.fromisoformat(points[-1]['date'])
+    days = max(1, (end - begin).days)
+    maximum = max(500, math.ceil(max(p['stars'] for p in points) / 500) * 500)
+    tick_step = 10 ** max(0, int(math.log10(maximum)))
+    xy = lambda day, n: (left + (right-left)*(date.fromisoformat(day)-begin).days / days,
+                         bottom-(bottom-top)*n/maximum)
+    for value in range(0, maximum+1, tick_step):
+        y = xy(points[0]['date'], value)[1]
+        parts.append(f'<path d="M{left} {y:.2f} H{right}" stroke="{t["grid"]}" stroke-dasharray="4 6"/>')
+        parts.append(txt(left-16, round(y+4, 2), f'{value / 1000:g}k' if value >= 1000 else str(value), 12, t['muted'], anchor='end'))
+    dates = sorted({begin + timedelta(days=round((end-begin).days*i/5)) for i in range(6)})
+    ticks = [(d.isoformat(), d.strftime('%b %Y') if days > 90 else d.strftime('%b %d')) for d in dates]
+    for day, label in ticks:
+        x = xy(day, 0)[0]
+        parts.append(f'<path d="M{x:.2f} {top} V{bottom}" stroke="{t["grid"]}" stroke-opacity=".65"/>')
+        anchor = 'start' if day == points[0]['date'] else 'end' if day == snapshot else 'middle'
+        parts.append(txt(round(x, 2), 466, label, 12, t['muted'], anchor=anchor))
+    coords = [xy(p['date'], p['stars']) for p in points]
+    line = 'M' + ' L'.join(f'{x:.2f} {y:.2f}' for x, y in coords)
+    area = line + f' L{coords[-1][0]:.2f} {bottom} L{left} {bottom} Z'
+    parts.extend([
+        f'<path d="{area}" fill="url(#area)"/>',
+        f'<path d="{line}" fill="none" stroke="url(#accent)" stroke-width="3" '
+        'stroke-linecap="round" stroke-linejoin="round"/>',
+        f'<path d="M{left} {bottom} H{right}" stroke="{t["edge"]}"/>',
+    ])
+    x, y = coords[-1]
+    parts.extend([
+        f'<circle cx="{x}" cy="{y}" r="10" fill="{t["copper"]}" opacity=".12"/>',
+        f'<circle cx="{x}" cy="{y}" r="5" fill="{t["copper"]}" stroke="{t["bg"]}" stroke-width="2"/>',
+        f'<path d="M40 491 H960" stroke="{t["edge"]}"/>',
+        txt(40, 516, f'{begin:%b %Y} — {end:%b %Y}'.upper(), 10, t['muted'], 500, mono=True, spacing=.7),
+        txt(959, 516, f'SNAPSHOT {snapshot}', 10, t['muted'], 500, 'end', mono=True, spacing=.6),
+        txt(40, 539, provenance, 11, t['muted']),
+        '</svg>',
+    ])
+    return ''.join(parts)
@@ -0,0 +1 @@
+PyYAML==6.0.3
@@ -0,0 +1,178 @@
+# Copyright (c) 2026 Feng Ruohang
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+"""Regression checks for historical accuracy, contributor scope and SVG safety."""
+
+import base64
+import json
+from pathlib import Path
+import tempfile
+import unittest
+from unittest.mock import patch
+from urllib.error import URLError
+import xml.etree.ElementTree as ET
+
+import render
+import update
+
+NS = {'s': 'http://www.w3.org/2000/svg'}
+PNG = base64.b64decode('iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+a7mgAAAAASUVORK5CYII=')
+
+
+def person(handle='Alice', group='reports', featured=False):
+    return {'handle': handle, 'group': group, 'featured': featured,
+            'what': 'A reviewed contribution', 'firstContribution': '2026-09-01'}
+
+
+def history():
+    return {'repository': 'pgsty/silo',
+            'bootstrap': {'through': '2026-09-15', 'reconstructed': True},
+            'points': [{'date': '2026-09-14', 'stars': 100}, {'date': '2026-09-15', 'stars': 105}]}
+
+
+class HistoryTests(unittest.TestCase):
+    def test_new_day_preserves_old_counts_and_unstars(self):
+        before = history()
+        after = update.update_history(before, '2026-09-16', 103)
+        self.assertEqual(after['points'][:-1], before['points'])
+        self.assertEqual(after['points'][-1], {'date': '2026-09-16', 'stars': 103})
+        self.assertEqual(before, history())
+
+    def test_same_day_rerun_replaces_instead_of_appending(self):
+        first = update.update_history(history(), '2026-09-15', 107)
+        self.assertEqual(len(first['points']), 2)
+        self.assertEqual(first, update.update_history(first, '2026-09-15', 107))
+
+    def test_missing_days_are_not_invented(self):
+        result = update.update_history(history(), '2026-09-18', 106)
+        self.assertEqual([p['date'] for p in result['points']], ['2026-09-14', '2026-09-15', '2026-09-18'])
+
+    def test_rejects_wrong_repository_and_corrupt_history(self):
+        cases = []
+        wrong = history(); wrong['repository'] = 'someone/else'; cases.append(wrong)
+        duplicate = history(); duplicate['points'].append(duplicate['points'][-1]); cases.append(duplicate)
+        unordered = history(); unordered['points'].reverse(); cases.append(unordered)
+        negative = history(); negative['points'][0]['stars'] = -1; cases.append(negative)
+        future = history(); future['points'][-1]['date'] = '2026-09-20'; cases.append(future)
+        for case in cases:
+            with self.subTest(case=case), self.assertRaises(ValueError):
+                update.update_history(case, '2026-09-16', 100)
+
+    def test_first_run_has_no_fabricated_history(self):
+        result = update.update_history(None, '2026-09-16', 10)
+        self.assertFalse(result['bootstrap']['reconstructed'])
+        self.assertEqual(result['points'], [{'date': '2026-09-16', 'stars': 10}])
+
+
+class ContributorTests(unittest.TestCase):
+    def test_retries_truncated_json_before_using_it(self):
+        with patch('update.request', side_effect=[b'{"partial":', b'{"ok":true}']), patch('update.time.sleep'):
+            self.assertEqual(update.GitHub('').get('repos/pgsty/silo'), {'ok': True})
+
+    def test_paginates_past_one_full_page(self):
+        class API(update.GitHub):
+            def __init__(self): self.calls = []
+            def get(self, path):
+                self.calls.append(path)
+                return list(range(100)) if 'page=1&' in path else [100]
+        api = API()
+        self.assertEqual(len(list(api.issues('pgsty/silo'))), 101)
+        self.assertIn('state=all', api.calls[0])
+        self.assertIn('page=2&', api.calls[1])
+
+    def test_bots_deduplication_unmerged_work_and_reviewed_credit(self):
+        def issue(login, kind='issue', user_type='User'):
+            item = {'user': {'login': login, 'type': user_type, 'avatar_url': ''}, 'created_at': '2026-09-02T00:00:00Z'}
+            if kind != 'issue': item['pull_request'] = {'merged_at': None if kind == 'open' else '2026-09-03T00:00:00Z'}
+            return item
+        class API:
+            def issues(self, _repo):
+                return [issue('alice'), issue('Bob', 'open'), issue('Bob', 'merged'),
+                        issue('Carol', 'open'), issue('Copilot'), issue('robot', user_type='Bot')]
+        curated = {'repositories': ['pgsty/silo', 'pgsty/mc'], 'bots': ['Copilot'],
+                   'people': [person('Alice', featured=True), person('Reporter'), person('Copilot')]}
+        result = update.collect_people(API(), curated)
+        self.assertEqual({p['handle'] for p in result}, {'Alice', 'Bob', 'Carol', 'Reporter'})
+        self.assertEqual(result[0]['handle'], 'Bob')
+        self.assertEqual(result[1]['handle'], 'Carol')
+        self.assertTrue(next(p for p in result if p['handle'] == 'Alice')['featured'])
+        self.assertEqual(next(p for p in result if p['handle'] == 'Bob')['group'], 'code')
+        self.assertFalse(next(p for p in result if p['handle'] == 'Carol')['featured'])
+
+    def test_newer_reviewed_preview_survives_until_site_catches_up(self):
+        remote = {'updated': '2026-09-16T03:00:00+00:00'}
+        cached = {'updated': '2026-09-16T04:00:00+00:00'}
+        self.assertIs(update.select_curated(remote, cached), cached)
+        newer = {'updated': '2026-09-17T03:00:00+00:00'}
+        self.assertIs(update.select_curated(newer, cached), newer)
+
+    def test_avatar_failure_reuses_raster_cache(self):
+        previous = {**person(), 'avatarDataUrl': update.raster_data_url(PNG)}
+        with patch('update.request', side_effect=URLError('unavailable')):
+            result = update.add_avatars(None, [{**person(), 'avatarUrl': 'https://avatars.githubusercontent.com/u/1'}], [previous])
+        self.assertEqual(result[0]['avatarDataUrl'], previous['avatarDataUrl'])
+        with self.assertRaises(ValueError): update.raster_data_url(b'<svg onload="bad()"/>')
+        with self.assertRaises(ValueError): update.cached_avatar({'avatarDataUrl': 'data:image/svg+xml;base64,PHN2Zy8+'})
+
+    def test_fetch_failure_leaves_published_assets_untouched(self):
+        class API:
+            def get(self, path):
+                if path == 'repos/pgsty/silo': return {'full_name': 'pgsty/silo', 'stargazers_count': 106}
+                raise URLError('roster unavailable')
+        with tempfile.TemporaryDirectory() as directory:
+            out = Path(directory)
+            original = json.dumps(history())
+            (out / 'history.json').write_text(original)
+            (out / 'contributors-light.svg').write_text('previous image')
+            with self.assertRaises(URLError): update.refresh(out, API(), Path('.'))
+            self.assertEqual((out / 'history.json').read_text(), original)
+            self.assertEqual((out / 'contributors-light.svg').read_text(), 'previous image')
+
+
+class RenderTests(unittest.TestCase):
+    def test_real_emblem_generates_clean_xml(self):
+        emblem = render.read_emblem(Path(__file__).resolve().parents[2] / '.github/silo.svg')
+        svg = render.contributors('light', [person()], '2026-09-16', emblem)
+        ET.fromstring(svg)
+        self.assertTrue(all(line == line.rstrip() for line in svg.splitlines()))
+
+    def test_all_avatars_fit_when_the_roster_grows(self):
+        people = [{**person(f'person-{i}'), 'avatarDataUrl': update.raster_data_url(PNG)} for i in range(151)]
+        for theme in ('light', 'dark'):
+            root = ET.fromstring(render.contributors(theme, people, '2026-09-16', ''))
+            images = root.findall('.//s:image', NS)
+            self.assertEqual(len(images), 151)
+            footer = float(root.attrib['height']) - 42
+            self.assertTrue(all(float(i.attrib['y']) + float(i.attrib['height']) < footer for i in images))
+            self.assertTrue(all(i.attrib['href'].startswith('data:image/png;base64,') for i in images))
+
+    def test_untrusted_text_is_escaped(self):
+        data = [{**person(), 'what': '<script>alert("x")</script> & contributions'}]
+        svg = render.contributors('light', data, '2026-09-16', '')
+        root = ET.fromstring(svg)
+        self.assertEqual(root.findall('.//s:script', NS), [])
+        self.assertIn('&lt;script&gt;', svg)
+
+    def test_single_point_and_decreasing_star_history_render(self):
+        for data in (update.update_history(None, '2026-09-16', 0), update.update_history(history(), '2026-09-16', 90)):
+            for theme in ('light', 'dark'):
+                svg = render.stars(theme, data, '2026-09-16', '')
+                ET.fromstring(svg)
+                self.assertNotIn('nan', svg.lower())
+                self.assertNotIn('inf', svg.lower())
+
+
+if __name__ == '__main__':
+    unittest.main()
@@ -0,0 +1,295 @@
+#!/usr/bin/env python3
+# Copyright (c) 2026 Feng Ruohang
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+"""Refresh the generated-asset checkout; publishing is handled by the workflow."""
+
+import argparse
+import base64
+from concurrent.futures import ThreadPoolExecutor
+from datetime import date, datetime, timezone
+import json
+from http.client import IncompleteRead
+import os
+from pathlib import Path
+import re
+import sys
+import time
+from urllib.error import HTTPError, URLError
+from urllib.parse import urlparse
+from urllib.request import Request, urlopen
+import xml.etree.ElementTree as ET
+
+import yaml
+
+import render
+
+REPOSITORY = 'pgsty/silo'
+SOURCE = 'repos/pgsty/silo.pgsty.com/contents/data/home/contributors.yaml?ref=main'
+GROUPS = ('code', 'proposed', 'reports')
+HANDLE = re.compile(r'[A-Za-z0-9][A-Za-z0-9-]{0,38}\Z')
+
+
+def request(url, token='', limit=8 * 1024 * 1024):
+    headers = {'User-Agent': 'silo-repository-cards', 'Accept': 'application/vnd.github+json'}
+    if urlparse(url).netloc == 'api.github.com':
+        headers['X-GitHub-Api-Version'] = '2022-11-28'
+        if token:
+            headers['Authorization'] = f'Bearer {token}'
+    for attempt in range(3):
+        try:
+            with urlopen(Request(url, headers=headers), timeout=25) as response:
+                data = response.read(limit + 1)
+                if len(data) > limit:
+                    raise ValueError('Response exceeds the size limit')
+                expected = response.headers.get('Content-Length')
+                if expected is not None and len(data) != int(expected):
+                    raise URLError('Incomplete response body')
+                return data
+        except HTTPError as exc:
+            if exc.code < 500 or attempt == 2:
+                raise
+        except (URLError, TimeoutError, IncompleteRead):
+            if attempt == 2:
+                raise
+        time.sleep(attempt + 1)
+
+
+class GitHub:
+    def __init__(self, token):
+        self.token = token
+
+    def get(self, path):
+        for attempt in range(3):
+            try:
+                return json.loads(request('https://api.github.com/' + path, self.token))
+            except (json.JSONDecodeError, UnicodeDecodeError) as exc:
+                if attempt == 2:
+                    raise ValueError(f'Incomplete or invalid GitHub JSON: {path}') from exc
+                time.sleep(attempt + 1)
+
+    def issues(self, repository):
+        page = 1
+        while True:
+            batch = self.get(f'repos/{repository}/issues?state=all&per_page=100&page={page}&sort=created&direction=asc')
+            if not isinstance(batch, list):
+                raise ValueError(f'Invalid issues response for {repository}')
+            yield from batch
+            if len(batch) < 100:
+                return
+            page += 1
+
+
+def curated_snapshot(data, revision):
+    updated = str(data['updated'])
+    datetime.fromisoformat(updated)
+    repositories = [item['repo'] for item in data['repositories']]
+    if not repositories or any(not re.fullmatch(r'pgsty/[A-Za-z0-9_.-]+', repo) for repo in repositories):
+        raise ValueError('Invalid contributor repository scope')
+    people = []
+    for group in GROUPS:
+        for entry in data[group]:
+            if not HANDLE.fullmatch(entry['handle']):
+                raise ValueError('Invalid GitHub contributor handle')
+            people.append({
+                'handle': entry['handle'], 'group': group,
+                'featured': bool(entry.get('featured')), 'what': entry['what'],
+                'firstContribution': str(entry.get('firstContribution', '9999-12-31')),
+            })
+    if not people or len({p['handle'].lower() for p in people}) != len(people):
+        raise ValueError('Empty or duplicate contributor roster')
+    return {'updated': updated, 'revision': revision, 'repositories': repositories,
+            'bots': data.get('bots', ['Copilot', 'dependabot[bot]']), 'people': people}
+
+
+def select_curated(remote, cached):
+    # The initial, approved preview can contain reviewed credit not published by
+    # the companion site yet. Keep that newer snapshot until the site catches up.
+    if cached and datetime.fromisoformat(cached['updated']) > datetime.fromisoformat(remote['updated']):
+        return cached
+    return remote
+
+
+def collect_people(api, curated):
+    bots = {name.lower() for name in curated['bots']}
+    people = {p['handle'].lower(): dict(p) for p in curated['people']
+              if p['handle'].lower() not in bots and not p['handle'].lower().endswith('[bot]')}
+    order = {p['handle'].lower(): index for index, p in enumerate(curated['people'])}
+    for repository in curated['repositories']:
+        print(f'Reading issue and PR authors: {repository}', flush=True)
+        for issue in api.issues(repository):
+            user = issue.get('user') or {}
+            handle = user.get('login', '')
+            key = handle.lower()
+            if user.get('type') != 'User' or key in bots or key.endswith('[bot]'):
+                continue
+            if not HANDLE.fullmatch(handle):
+                raise ValueError('Invalid issue author')
+            pr = issue.get('pull_request')
+            group = 'code' if pr and pr.get('merged_at') else 'proposed' if pr else 'reports'
+            first = issue['created_at'][:10]
+            date.fromisoformat(first)
+            person = people.setdefault(key, {
+                'handle': handle, 'group': group, 'featured': False,
+                'what': 'Contributed an issue or pull request to SILO and related projects',
+                'firstContribution': first,
+            })
+            person['avatarUrl'] = user.get('avatar_url', '')
+            person['firstContribution'] = min(person['firstContribution'], first)
+            if GROUPS.index(group) < GROUPS.index(person['group']):
+                person['group'] = group
+    if not people:
+        raise ValueError('No human contributors were collected')
+    return sorted(people.values(), key=lambda p: (
+        GROUPS.index(p['group']), not p['featured'],
+        order.get(p['handle'].lower(), len(order)), p['firstContribution'], p['handle'].lower()))
+
+
+def raster_data_url(data):
+    if data.startswith(b'\x89PNG\r\n\x1a\n'):
+        mime = 'image/png'
+    elif data.startswith(b'\xff\xd8\xff'):
+        mime = 'image/jpeg'
+    elif data.startswith((b'GIF87a', b'GIF89a')):
+        mime = 'image/gif'
+    elif data[:4] == b'RIFF' and data[8:12] == b'WEBP':
+        mime = 'image/webp'
+    else:
+        raise ValueError('Avatar is not a raster image')
+    return f'data:{mime};base64,' + base64.b64encode(data).decode('ascii')
+
+
+def cached_avatar(person):
+    value = person.get('avatarDataUrl', '')
+    if not value:
+        return ''
+    prefix, encoded = value.split(',', 1)
+    if prefix not in ('data:image/png;base64', 'data:image/jpeg;base64', 'data:image/gif;base64', 'data:image/webp;base64'):
+        raise ValueError('Invalid cached avatar format')
+    raw = base64.b64decode(encoded, validate=True)
+    if len(raw) > 512 * 1024 or raster_data_url(raw) != value:
+        raise ValueError('Invalid cached avatar')
+    return value
+
+
+def add_avatars(api, people, previous):
+    cached = {p['handle'].lower(): cached_avatar(p) for p in previous}
+
+    def update(person):
+        person = dict(person)
+        try:
+            url = person.pop('avatarUrl', '') or api.get('users/' + person['handle'])['avatar_url']
+            parsed = urlparse(url)
+            if parsed.scheme != 'https' or parsed.netloc != 'avatars.githubusercontent.com':
+                raise ValueError('Unexpected avatar host')
+            data = request(url + ('&' if '?' in url else '?') + 's=96', limit=512 * 1024)
+            person['avatarDataUrl'] = raster_data_url(data)
+        except (HTTPError, URLError, TimeoutError, IncompleteRead, ValueError, KeyError) as exc:
+            person.pop('avatarUrl', None)
+            person['avatarDataUrl'] = cached.get(person['handle'].lower(), '')
+            print(f'Avatar fallback for @{person["handle"]}: {type(exc).__name__}', file=sys.stderr)
+        return person
+
+    with ThreadPoolExecutor(max_workers=6) as pool:
+        return list(pool.map(update, people))
+
+
+def update_history(history, day, stars):
+    date.fromisoformat(day)
+    if type(stars) is not int or stars < 0:
+        raise ValueError('Invalid repository star count')
+    if history is None:
+        history = {'repository': REPOSITORY, 'bootstrap': {'through': day, 'reconstructed': False}, 'points': []}
+    if history['repository'] != REPOSITORY:
+        raise ValueError('Star history belongs to a different repository')
+    date.fromisoformat(history['bootstrap']['through'])
+    dates = []
+    for point in history['points']:
+        date.fromisoformat(point['date'])
+        if type(point['stars']) is not int or point['stars'] < 0:
+            raise ValueError('Invalid historical star count')
+        dates.append(point['date'])
+    if dates != sorted(set(dates)) or any(d > day for d in dates):
+        raise ValueError('History contains duplicate, unordered, or future dates')
+    # Replace today's observation, preserve previous days, and allow unstars.
+    points = [dict(p) for p in history['points'] if p['date'] != day]
+    points.append({'date': day, 'stars': stars})
+    return {**history, 'points': points}
+
+
+def read_json(path, default=None):
+    return json.loads(path.read_text()) if path.exists() else default
+
+
+def refresh(output, api, source_root):
+    day = datetime.now(timezone.utc).date().isoformat()
+    metadata = api.get('repos/' + REPOSITORY)
+    if metadata['full_name'].lower() != REPOSITORY:
+        raise ValueError('Unexpected repository metadata')
+    history = update_history(read_json(output / 'history.json'), day, metadata['stargazers_count'])
+    source = api.get(SOURCE)
+    reviewed = yaml.safe_load(base64.b64decode(source['content'], validate=False))
+    curated = select_curated(curated_snapshot(reviewed, source['sha']), read_json(output / 'curated.json'))
+    people = collect_people(api, curated)
+    previous = read_json(output / 'contributors.json', {}).get('people', [])
+    people = add_avatars(api, people, previous)
+    emblem = render.read_emblem(source_root / '.github/silo.svg')
+    payloads = {}
+    for theme in ('light', 'dark'):
+        payloads[f'contributors-{theme}.svg'] = render.contributors(theme, people, day, emblem) + '\n'
+        payloads[f'star-history-{theme}.svg'] = render.stars(theme, history, day, emblem) + '\n'
+    for svg in payloads.values():
+        ET.fromstring(svg)
+    for name, data in {
+        'history.json': history,
+        'curated.json': curated,
+        'contributors.json': {'repository': REPOSITORY, 'updated': day, 'people': people},
+    }.items():
+        payloads[name] = json.dumps(data, indent=2, ensure_ascii=False) + '\n'
+    payloads['README.md'] = f'''# SILO repository cards
+
+Generated by [Repository Cards](https://github.com/pgsty/silo/actions/workflows/repository-cards.yml)
+at 00:00 UTC daily (08:00 Asia/Shanghai). GitHub may queue scheduled runs.
+
+Snapshot: {day}. {metadata['stargazers_count']:,} stars; {len(people)} community contributors.
+
+- `contributors-light.svg` / `contributors-dark.svg`: human issue and PR authors across the SILO project scope, plus reviewed acknowledgements. Bots are excluded. Gold rings follow the reviewed companion-site roster; new authors are collected automatically.
+- `star-history-light.svg` / `star-history-dark.svg`: initial history reconstructed from the then-current stargazers; later points are daily observed totals, including decreases. Missing days are not fabricated.
+- `curated.json`: a cache of reviewed contributor credit from `pgsty/silo.pgsty.com/data/home/contributors.yaml`. The approved initial preview may be newer than the published site; a newer reviewed snapshot is retained until the site catches up.
+- `contributors.json`: generated contributor data and embedded raster avatars. Failed avatar refreshes use the previous image, or an initial when no image is available.
+- `history.json`: persistent daily totals. Keep this file when regenerating images.
+
+The SVGs are self-contained. Source and instructions live on the default branch;
+this branch contains generated assets only. Do not merge it into `main`.
+'''
+    # Collect and validate everything before touching the publication checkout.
+    output.mkdir(parents=True, exist_ok=True)
+    for filename, text in payloads.items():
+        (output / filename).write_text(text)
+    print(f'{day}: {len(people)} contributors; {metadata["stargazers_count"]:,} stars; {len(history["points"])} history points')
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--output', type=Path, required=True)
+    args = parser.parse_args()
+    configured = os.environ.get('GITHUB_REPOSITORY', REPOSITORY)
+    if configured.lower() != REPOSITORY:
+        raise SystemExit('This workflow is scoped to pgsty/silo')
+    refresh(args.output, GitHub(os.environ.get('GH_TOKEN', '')), Path(__file__).resolve().parents[2])
+
+
+if __name__ == '__main__':
+    main()
@@ -113,7 +113,7 @@ helm_run template my-release "${new_chart}" \
 go run ./buildscripts/helm-migration-guard "${old_render}" "${new_render}"

 helm_run package "${new_chart}" --destination "${output_dir}" >/dev/null
-test -s "${work_dir}/silo-7.0.2.tgz"
+test -s "${work_dir}/silo-7.0.3.tgz"
 if find "${work_dir}" -maxdepth 1 -type f -name 'minio-*.tgz' | grep -q .; then
 	echo "Helm packaging emitted a legacy MinIO chart name" >&2
 	exit 1
@@ -163,6 +163,7 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {

 		// StorageInfo operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/storageinfo").HandlerFunc(adminMiddleware(adminAPI.StorageInfoHandler, traceAllFlag))
+		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/multipart-preflight").HandlerFunc(adminMiddleware(adminAPI.MultipartPreflightHandler, traceAllFlag))
 		// DataUsageInfo operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/datausageinfo").HandlerFunc(adminMiddleware(adminAPI.DataUsageInfoHandler, traceAllFlag))
 		// Metrics operation
@@ -450,6 +450,9 @@ const (
 	ErrAdminNoSecretKey

 	ErrIAMNotInitialized
+	ErrMultipartListingLegacy
+	ErrMultipartListingIdentity
+	ErrSlowDown

 	apiErrCodeEnd // This is used only for the testing code
 )
@@ -1336,6 +1339,21 @@ var errorCodes = errorCodeMap{
 		Description:    "IAM sub-system not initialized yet, please try again.",
 		HTTPStatusCode: http.StatusServiceUnavailable,
 	},
+	ErrMultipartListingLegacy: {
+		Code:           "MultipartListingNotReady",
+		Description:    "Legacy multipart uploads prevent a complete listing. Upgrade all writers, drain old uploads and run the multipart preflight check.",
+		HTTPStatusCode: http.StatusServiceUnavailable,
+	},
+	ErrMultipartListingIdentity: {
+		Code:           "MultipartListingMetadataInvalid",
+		Description:    "Multipart upload metadata is inconsistent. Run the multipart preflight check to locate the affected storage set.",
+		HTTPStatusCode: http.StatusServiceUnavailable,
+	},
+	ErrSlowDown: {
+		Code:           "SlowDown",
+		Description:    "Please reduce your request rate",
+		HTTPStatusCode: http.StatusServiceUnavailable,
+	},
 	ErrBucketMetadataNotInitialized: {
 		Code:           "XMinioBucketMetadataNotInitialized",
 		Description:    "Bucket metadata not initialized yet, please try again.",
@@ -2173,6 +2191,10 @@ func toAPIErrorCode(ctx context.Context, err error) (apiErr APIErrorCode) {
 	err = unwrapAll(err)

 	switch err {
+	case errMultipartListingLegacy:
+		apiErr = ErrMultipartListingLegacy
+	case errMultipartListingIdentity:
+		apiErr = ErrMultipartListingIdentity
 	case errCompleteMultipartChecksumMismatch, errCompleteMultipartChecksumTypeMismatch:
 		apiErr = ErrBadDigest
 	case errMissingPartChecksum:
@@ -2303,6 +2325,8 @@ func toAPIErrorCode(ctx context.Context, err error) (apiErr APIErrorCode) {
 	}

 	switch err.(type) {
+	case SlowDown:
+		apiErr = ErrSlowDown
 	case StorageFull:
 		apiErr = ErrStorageFull
 	case hash.BadDigest:
@@ -41,7 +41,7 @@ import (
 const (
 	maxObjectList  = 1000  // Limit number of objects in a listObjectsResponse/listObjectsVersionsResponse.
 	maxDeleteList  = 1000  // Limit number of objects deleted in a delete call.
-	maxUploadsList = 10000 // Limit number of uploads in a listUploadsResponse.
+	maxUploadsList = 1000  // Limit number of uploads in a listUploadsResponse.
 	maxPartsList   = 10000 // Limit number of parts in a listPartsResponse.
 )

@@ -277,14 +277,6 @@ func (api objectAPIHandlers) ListMultipartUploadsHandler(w http.ResponseWriter,
 		return
 	}

-	if keyMarker != "" {
-		// Marker not common with prefix is not implemented.
-		if !HasPrefix(keyMarker, prefix) {
-			writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrNotImplemented), r.URL)
-			return
-		}
-	}
-
 	listMultipartsInfo, err := objectAPI.ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
@@ -425,7 +425,7 @@ func testListMultipartUploadsHandler(obj ObjectLayer, instanceType, bucketName s
 			shouldPass:         true,
 		},
 		// Test case - 4.
-		// Setting Invalid prefix and marker combination.
+		// A key marker outside the prefix is valid and produces an empty page.
 		{
 			bucket:             bucketName,
 			prefix:             "asia",
@@ -435,8 +435,8 @@ func testListMultipartUploadsHandler(obj ObjectLayer, instanceType, bucketName s
 			maxUploads:         "0",
 			accessKey:          credentials.AccessKey,
 			secretKey:          credentials.SecretKey,
-			expectedRespStatus: http.StatusNotImplemented,
-			shouldPass:         false,
+			expectedRespStatus: http.StatusOK,
+			shouldPass:         true,
 		},
 		// Test case - 5.
 		// Invalid upload id and marker combination.
@@ -504,6 +504,35 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	ctx = lkctx.Context()
 	defer lk.Unlock(lkctx)

+	if !isPurge && dobj.DeleteMarkerVersionID != "" {
+		// A creation task carries the marker as it looked when it was queued
+		// by the DELETE handler, a GET/HEAD/LIST heal, the scanner or MRF.
+		// While it waited for this lock, another frontend may have purged the
+		// marker and replicated that purge. The targets no longer hold the
+		// marker, so the queued creation would recreate it from the stale
+		// snapshot. Confirm the source version under the lock first.
+		switch deleteMarkerCreationState(ctx, objectAPI, dobj) {
+		case creationStale:
+			return replicatedInfos{}
+		case creationUnverified:
+			dobj.RetryCount++
+			globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
+			sendEvent(eventArgs{
+				BucketName: bucket,
+				Object: ObjectInfo{
+					Bucket:       bucket,
+					Name:         dobj.ObjectName,
+					VersionID:    versionID,
+					DeleteMarker: dobj.DeleteMarker,
+				},
+				UserAgent: "Internal: [Replication]",
+				Host:      globalLocalNodeName,
+				EventName: event.ObjectReplicationNotTracked,
+			})
+			return replicatedInfos{}
+		}
+	}
+
 	rinfos := replicatedInfos{Targets: make([]replicatedTargetInfo, 0, len(dsc.targetsMap))}
 	var wg sync.WaitGroup
 	var mu sync.Mutex
@@ -1955,6 +1984,41 @@ func (di DeletedObjectReplicationInfo) isVersionPurge() bool {
 	return di.VersionID != "" || di.DeleteMarkerVersionID != "" && !di.VersionPurgeStatus().Empty()
 }

+// creationState is the outcome of re-reading a queued delete-marker creation
+// against the source.
+type creationState int
+
+const (
+	creationCurrent creationState = iota
+	creationStale
+	creationUnverified
+)
+
+// deleteMarkerCreationState re-reads the marker a queued creation task refers
+// to. The version being absent, no longer a marker, or under a version purge
+// makes the creation stale. A failed read is not absence: the caller retries
+// later instead of guessing.
+func deleteMarkerCreationState(ctx context.Context, objectAPI ObjectLayer, dobj DeletedObjectReplicationInfo) creationState {
+	oi, err := objectAPI.GetObjectInfo(ctx, dobj.Bucket, dobj.ObjectName, ObjectOptions{
+		VersionID:        dobj.DeleteMarkerVersionID,
+		Versioned:        globalBucketVersioningSys.PrefixEnabled(dobj.Bucket, dobj.ObjectName),
+		VersionSuspended: globalBucketVersioningSys.Suspended(dobj.Bucket),
+	})
+	switch {
+	case isErrObjectNotFound(err), isErrVersionNotFound(err):
+		return creationStale
+	case err != nil && !isErrMethodNotAllowed(err):
+		return creationUnverified
+	}
+	if !oi.DeleteMarker || oi.VersionID != dobj.DeleteMarkerVersionID {
+		return creationStale
+	}
+	if !oi.VersionPurgeStatus.Empty() || oi.VersionPurgeStatusInternal != "" {
+		return creationStale
+	}
+	return creationCurrent
+}
+
 // Purge metadata uses COMPLETE; operation statistics and audit use COMPLETED.
 func purgeReplicationStatus(status VersionPurgeStatusType) replication.StatusType {
 	if replication.StatusType(status) == replication.CompletedLegacy {
@@ -0,0 +1,109 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/minio/minio/internal/auth"
+	"github.com/minio/minio/internal/bucket/replication"
+	xhttp "github.com/minio/minio/internal/http"
+)
+
+// An exact-version DELETE without a replication decision physically purges
+// the version. Its response must describe the stored version: a data version
+// is not a delete marker even while pending-purge metadata makes the lookup
+// expose it as deleted, and a stored marker stays a marker.
+func TestDeleteMarkerPurgeResponseIdentity(t *testing.T) {
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, endpoints: []string{"DeleteObject"}, objAPITest: func(obj ObjectLayer, backend, bucket string, router http.Handler, creds auth.Credentials, t *testing.T) {
+		defer replicationTestCapacity(obj)()
+		ctx := t.Context()
+		if _, err := globalBucketMetadataSys.Update(ctx, bucket, bucketVersioningConfig, enabledBucketVersioningConfig); err != nil {
+			t.Fatal(err)
+		}
+		// Seed the purge state the DELETE handler would record for a target.
+		arn := "arn:minio:replication::" + mustGetUUID() + ":bucket"
+		pendingPurge := ReplicationState{
+			VersionPurgeStatusInternal: arn + "=PENDING;",
+			PurgeTargets:               map[string]VersionPurgeStatusType{arn: replication.VersionPurgePending},
+		}
+		for _, tc := range []struct {
+			name           string
+			marker, purged bool
+		}{
+			{name: "data"},
+			{name: "data-pending-purge", purged: true},
+			{name: "marker", marker: true},
+			{name: "marker-pending-purge", marker: true, purged: true},
+		} {
+			name := "purge-response-" + mustGetUUID()
+			oi, err := obj.PutObject(ctx, bucket, name, mustGetPutObjReader(t, bytes.NewReader([]byte("data")), 4, "", ""), ObjectOptions{Versioned: true})
+			if err != nil {
+				t.Fatal(err)
+			}
+			version := oi.VersionID
+			if tc.marker {
+				opts := ObjectOptions{Versioned: true, VersionID: mustGetUUID(), DeleteMarker: true, MTime: UTCNow(), ReplicationRequest: true}
+				opts.SetReplicaStatus(replication.Replica)
+				if _, err := obj.DeleteObject(ctx, bucket, name, opts); err != nil {
+					t.Fatal(err)
+				}
+				version = opts.VersionID
+			}
+			if tc.purged {
+				if _, err := obj.DeleteObject(ctx, bucket, name, ObjectOptions{Versioned: true, VersionID: version, DeleteReplication: pendingPurge}); err != nil {
+					t.Fatal(err)
+				}
+			}
+			before, err := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{Versioned: true, VersionID: version})
+			if tc.marker || tc.purged {
+				if !isErrMethodNotAllowed(err) || !before.DeleteMarker {
+					t.Fatalf("%s: seeded version not exposed as deleted: %+v %v", tc.name, before, err)
+				}
+			} else if err != nil || before.DeleteMarker {
+				t.Fatalf("%s: seeded data version: %+v %v", tc.name, before, err)
+			}
+			if tc.purged && before.VersionPurgeStatus != replication.VersionPurgePending {
+				t.Fatalf("%s: purge state not seeded: %+v", tc.name, before)
+			}
+			req, err := newTestSignedRequestV4(http.MethodDelete, "/"+bucket+"/"+name+"?versionId="+version, 0, nil, creds.AccessKey, creds.SecretKey, nil)
+			if err != nil {
+				t.Fatal(err)
+			}
+			rec := httptest.NewRecorder()
+			router.ServeHTTP(rec, req)
+			if rec.Code != http.StatusNoContent {
+				t.Fatalf("%s: status=%d: %s", tc.name, rec.Code, rec.Body.String())
+			}
+			// The handler sets these headers by map key, not canonical name.
+			if got := rec.Header()[xhttp.AmzVersionID]; len(got) != 1 || got[0] != version {
+				t.Fatalf("%s: response version %q, want %q", tc.name, got, version)
+			}
+			if got := len(rec.Header()[xhttp.AmzDeleteMarker]) == 1 && rec.Header()[xhttp.AmzDeleteMarker][0] == "true"; got != tc.marker {
+				t.Fatalf("%s: x-amz-delete-marker=%v for a stored marker=%v", tc.name, got, tc.marker)
+			}
+			if _, err := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{Versioned: true, VersionID: version}); !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+				t.Fatalf("%s: version not purged: %v", tc.name, err)
+			}
+			t.Logf("%s %s: purged with x-amz-delete-marker=%q", backend, tc.name, rec.Header()[xhttp.AmzDeleteMarker])
+		}
+	}})
+}
@@ -0,0 +1,524 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"maps"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio-go/v7"
+	"github.com/minio/minio/internal/bucket/replication"
+	xhttp "github.com/minio/minio/internal/http"
+	"github.com/minio/minio/internal/once"
+)
+
+func TestDeleteMarkerPurgeIntent(t *testing.T) {
+	version := mustGetUUID()
+	for _, tc := range []struct {
+		name string
+		edit func(*ObjectOptions)
+		want bool
+	}{
+		{"ordinary", func(*ObjectOptions) {}, true},
+		{"receiver", func(o *ObjectOptions) { o.ReplicationRequest = true; o.SetReplicaStatus(replication.Replica) }, true},
+		{"untrusted-replica", func(o *ObjectOptions) { o.SetReplicaStatus(replication.Replica) }, false},
+		{"complete-purge", func(o *ObjectOptions) {
+			o.DeleteReplication.VersionPurgeStatusInternal = string(replication.VersionPurgeComplete)
+		}, true},
+		{"pending-purge", func(o *ObjectOptions) { o.DeleteReplication.VersionPurgeStatusInternal = "PENDING" }, false},
+		{"failed-purge", func(o *ObjectOptions) { o.DeleteReplication.VersionPurgeStatusInternal = "FAILED" }, false},
+		{"unknown-purge", func(o *ObjectOptions) { o.DeleteReplication.VersionPurgeStatusInternal = "future-state" }, false},
+		{"pending-creation", func(o *ObjectOptions) { o.DeleteReplication.ReplicationStatusInternal = "PENDING" }, false},
+		{"failed-creation", func(o *ObjectOptions) { o.DeleteReplication.ReplicationStatusInternal = "FAILED" }, false},
+		{"completed-creation", func(o *ObjectOptions) { o.DeleteReplication.ReplicationStatusInternal = "COMPLETED" }, false},
+		{"create-marker", func(o *ObjectOptions) { o.DeleteMarker = true }, false},
+		{"empty", func(o *ObjectOptions) { o.VersionID = "" }, false},
+		{"null", func(o *ObjectOptions) { o.VersionID = nullVersionID }, false},
+		{"invalid", func(o *ObjectOptions) { o.VersionID = "invalid" }, false},
+		{"zero-uuid", func(o *ObjectOptions) { o.VersionID = emptyUUID }, false},
+		{"movement", func(o *ObjectOptions) { o.DataMovement = true }, false},
+		{"free-version", func(o *ObjectOptions) { o.InclFreeVersions = true }, false},
+		{"expiration", func(o *ObjectOptions) { o.Expiration.Expire = true }, false},
+		{"transition", func(o *ObjectOptions) { o.Transition.Status = "complete" }, false},
+		{"restored-expiration", func(o *ObjectOptions) { o.Transition.ExpireRestored = true }, false},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			o := ObjectOptions{VersionID: version, Versioned: true}
+			tc.edit(&o)
+			if got := o.isVersionPurge(); got != tc.want {
+				t.Fatalf("purge=%v want=%v", got, tc.want)
+			}
+		})
+	}
+}
+
+// Every other StorageAPI method panics through the nil embedding: a proof may
+// read metadata, but must never delete, heal, or ask a different storage API.
+type absenceProofDisk struct {
+	StorageAPI
+	t       *testing.T
+	err     error
+	reads   *atomic.Int32
+	deletes *atomic.Int32
+}
+
+func (d absenceProofDisk) ReadVersion(_ context.Context, _, _, _, _ string, opts ReadOptions) (FileInfo, error) {
+	if opts.ReadData || opts.Healing {
+		d.t.Error("absence proof requested data/healing")
+	}
+	d.reads.Add(1)
+	return FileInfo{}, d.err
+}
+
+func (d absenceProofDisk) DeleteVersion(_ context.Context, _, _ string, _ FileInfo, _ bool, _ DeleteOptions) error {
+	if d.deletes == nil {
+		d.t.Error("read-only absence proof attempted a deletion")
+	} else {
+		d.deletes.Add(1)
+	}
+	return d.err
+}
+
+func TestVersionPurgeAbsenceProof(t *testing.T) {
+	for _, tc := range []struct {
+		name        string
+		errs        []error
+		quorum, all bool
+	}{
+		{"all-absent", []error{errFileNotFound, errFileVersionNotFound, errFileNotFound, errFileVersionNotFound}, true, true},
+		{"majority-absent", []error{errFileNotFound, errFileVersionNotFound, errFileNotFound, nil}, true, false},
+		{"half-absent", []error{errFileNotFound, errFileVersionNotFound, errDiskNotFound, errDiskNotFound}, false, false},
+		{"corrupt", []error{errFileNotFound, errFileVersionNotFound, errFileCorrupt, errFileCorrupt}, false, false},
+		{"permission", []error{errFileNotFound, errFileVersionNotFound, errDiskAccessDenied, errVolumeAccessDenied}, false, false},
+		{"volume-missing", []error{errFileNotFound, errFileVersionNotFound, errVolumeNotFound, errVolumeNotFound}, false, false},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			var reads atomic.Int32
+			disks := make([]StorageAPI, len(tc.errs))
+			for i, err := range tc.errs {
+				disks[i] = absenceProofDisk{t: t, err: err, reads: &reads}
+			}
+			er := erasureObjects{getDisks: func() []StorageAPI { return disks }}
+			oldQueue := globalMRFState.opCh
+			globalMRFState.opCh = make(chan PartialOperation, 1)
+			defer func() { globalMRFState.opCh = oldQueue }()
+			all, err := er.confirmVersionAbsent(t.Context(), "bucket", "object", mustGetUUID())
+			if (err == nil) != tc.quorum || all != tc.all || reads.Load() != int32(len(disks)) || len(globalMRFState.opCh) != 0 {
+				t.Fatalf("proof all=%v error=%v reads=%d queued=%d", all, err, reads.Load(), len(globalMRFState.opCh))
+			}
+		})
+	}
+}
+
+func TestVersionPurgeAggregation(t *testing.T) {
+	for _, pure := range []bool{false, true} {
+		for _, fault := range []error{nil, errFileCorrupt, errDiskAccessDenied} {
+			t.Run(fmt.Sprintf("purge_%v_fault_%v", pure, fault), func(t *testing.T) {
+				var calls atomic.Int32
+				disks := make([]StorageAPI, 4)
+				for i, err := range []error{nil, errFileNotFound, errFileVersionNotFound, fault} {
+					disks[i] = absenceProofDisk{t: t, err: err, deletes: &calls}
+				}
+				er := erasureObjects{getDisks: func() []StorageAPI { return disks }}
+				err := er.deleteObjectVersion(t.Context(), "bucket", "object", FileInfo{VersionID: mustGetUUID()}, false, pure)
+				if (err == nil) != pure || calls.Load() != 4 {
+					t.Fatalf("aggregation error=%v calls=%d", err, calls.Load())
+				}
+			})
+		}
+	}
+}
+
+func TestDeleteMarkerPurgeOmittedPool(t *testing.T) {
+	z, bucket := consistencyPools(t)
+	defer replicationTestCapacity(z)()
+	const name = "partially-absent"
+	er := z.serverPools[1].getHashedSet(name)
+	_, opts := seedPurgeMarker(t, er, bucket, name, false)
+	opts.ReplicationRequest = false
+	opts.DeleteReplication = ReplicationState{}
+	// Pool 0 appears absent at read quorum, but half of it cannot be read.
+	missing := z.serverPools[0].getHashedSet(name)
+	original := missing.getDisks
+	disks := append([]StorageAPI(nil), original()...)
+	clear(disks[len(disks)/2:])
+	missing.getDisks = func() []StorageAPI { return disks }
+	defer func() { missing.getDisks = original }()
+	_, err := z.DeleteObject(t.Context(), bucket, name, opts)
+	if !isErrWriteQuorum(err) {
+		t.Errorf("omitted pool acknowledged deletion: %T %v", err, err)
+	}
+	if countPurgeMarkers(t, er.getDisks(), bucket, name, opts.VersionID) != 16 {
+		t.Fatal("mutated known copies before checking omitted pool")
+	}
+}
+
+func TestDeleteMarkerPurgeReceivingPool(t *testing.T) {
+	z, bucket := consistencyPools(t)
+	defer replicationTestCapacity(z)()
+	for _, duplicate := range []bool{false, true} {
+		t.Run(fmt.Sprintf("duplicate_%v", duplicate), func(t *testing.T) {
+			name := "receiver-" + mustGetUUID()
+			_, opts := seedPurgeMarker(t, z.serverPools[0].getHashedSet(name), bucket, name, false)
+			if duplicate {
+				create := opts
+				create.DeleteMarker = true
+				if _, err := z.serverPools[1].DeleteObject(t.Context(), bucket, name, create); err != nil {
+					t.Fatal(err)
+				}
+			} else {
+				// Latest-key routing selects pool 1, but the addressed marker is
+				// in pool 0. A replica purge addresses a version, not the latest.
+				putConsistencyObject(t, z, bucket, name, 1, "newer-data", ObjectOptions{Versioned: true})
+			}
+			if _, err := z.DeleteObject(t.Context(), bucket, name, opts); err != nil {
+				t.Errorf("replica purge did not find its addressed version: %v", err)
+			}
+			for i, pool := range z.serverPools {
+				if got := countPurgeMarkers(t, pool.getHashedSet(name).getDisks(), bucket, name, opts.VersionID); got != 0 {
+					t.Errorf("replica purge left %d marker copies in pool %d", got, i)
+				}
+			}
+		})
+	}
+}
+
+func TestDeleteMarkerPurgeCallbackIntent(t *testing.T) {
+	for _, loseCopies := range []bool{false, true} {
+		t.Run(fmt.Sprintf("missing_after_callback_%v", loseCopies), func(t *testing.T) {
+			z, bucket := consistencyPools(t)
+			defer replicationTestCapacity(z)()
+			const name = "callback-pending"
+			er := z.serverPools[0].getHashedSet(name)
+			_, opts := seedPurgeMarker(t, er, bucket, name, false)
+			original := er.getDisks
+			all := append([]StorageAPI(nil), original()...)
+			defer func() { er.getDisks = original }()
+			opts.ReplicationRequest = false
+			opts.DeleteReplication = ReplicationState{}
+			metadata, retention := 0, 0
+			opts.EvalRetentionBypassFn = func(oi ObjectInfo, err error) error {
+				retention++
+				if !oi.DeleteMarker || !isErrMethodNotAllowed(err) {
+					t.Fatalf("retention callback lost marker: %+v %v", oi, err)
+				}
+				return nil
+			}
+			opts.EvalMetadataFn = func(*ObjectInfo, error) (ReplicateDecision, error) {
+				metadata++
+				if loseCopies {
+					// Inject a disk-state change after preflight, before the set's
+					// metadata-update read. It returns NotFound at read quorum.
+					for _, disk := range all[:8] {
+						if err := disk.DeleteVersion(t.Context(), bucket, name, FileInfo{VersionID: opts.VersionID}, false, DeleteOptions{}); err != nil {
+							t.Fatal(err)
+						}
+					}
+					partial := make([]StorageAPI, len(all))
+					copy(partial, all[:8])
+					er.getDisks = func() []StorageAPI { return partial }
+				}
+				decision := ReplicateDecision{}
+				decision.Set(newReplicateTargetDecision("arn1", true, false))
+				return decision, nil
+			}
+			_, err := z.DeleteObject(t.Context(), bucket, name, opts)
+			if loseCopies {
+				if !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+					t.Errorf("pending update misclassified as physical purge: %T %v", err, err)
+				}
+			} else if err != nil {
+				t.Errorf("pending update failed: %v", err)
+			}
+			want := 16
+			if loseCopies {
+				want = 8
+			}
+			if got := countPurgeMarkers(t, all, bucket, name, opts.VersionID); got != want || metadata != 1 || retention != 1 {
+				t.Errorf("pending update: copies=%d want=%d metadata=%d retention=%d", got, want, metadata, retention)
+			}
+			if !loseCopies {
+				fi, err := all[0].ReadVersion(t.Context(), "", bucket, name, opts.VersionID, ReadOptions{})
+				if err != nil || fi.VersionPurgeStatus() != replication.VersionPurgePending {
+					t.Errorf("pending state not persisted: %+v %v", fi.ReplicationState, err)
+				}
+			}
+		})
+	}
+}
+
+func TestDeleteMarkerReceivingMetadataUpdate(t *testing.T) {
+	z, bucket := consistencyPools(t)
+	defer replicationTestCapacity(z)()
+	for _, state := range []VersionPurgeStatusType{replication.VersionPurgePending, replication.VersionPurgeFailed} {
+		t.Run(string(state), func(t *testing.T) {
+			name := "update-" + mustGetUUID()
+			_, opts := seedPurgeMarker(t, z.serverPools[0].getHashedSet(name), bucket, name, false)
+			create := opts
+			create.DeleteMarker = true
+			if _, err := z.serverPools[1].DeleteObject(t.Context(), bucket, name, create); err != nil {
+				t.Fatal(err)
+			}
+			opts.DeleteReplication = ReplicationState{VersionPurgeStatusInternal: string(state)}
+			if _, err := z.DeleteObject(t.Context(), bucket, name, opts); err != nil {
+				t.Fatal(err)
+			}
+			for i, pool := range z.serverPools {
+				if got := countPurgeMarkers(t, pool.getHashedSet(name).getDisks(), bucket, name, opts.VersionID); got != 16 {
+					t.Errorf("metadata update removed marker copies in pool %d: %d", i, got)
+				}
+			}
+		})
+	}
+}
+
+func TestVersionPurgeRetentionGate(t *testing.T) {
+	z, er, all, bucket := markerPurgeFixture(t, 4)
+	data, _ := seedPurgeMarker(t, er, bucket, "protected-data", true)
+	denied := errors.New("retention denied")
+	opts := ObjectOptions{Versioned: true, VersionID: data, EvalRetentionBypassFn: func(oi ObjectInfo, err error) error {
+		if err != nil || oi.VersionID != data || oi.DeleteMarker {
+			t.Errorf("wrong retained version: %+v error=%v", oi, err)
+		}
+		return denied
+	}}
+	_, err := z.DeleteObject(t.Context(), bucket, "protected-data", opts)
+	if !errors.Is(err, denied) {
+		t.Fatalf("retention gate: %v", err)
+	}
+	for _, disk := range all {
+		if _, err := disk.ReadVersion(t.Context(), "", bucket, "protected-data", data, ReadOptions{}); err != nil {
+			t.Errorf("protected version changed: %v", err)
+		}
+	}
+}
+
+func markerPurgeFixture(t *testing.T, n int) (*erasureServerPools, *erasureObjects, []StorageAPI, string) {
+	t.Helper()
+	ctx, cancel := context.WithCancel(t.Context())
+	obj, dirs, err := prepareErasure(ctx, n)
+	if err != nil {
+		cancel()
+		t.Fatal(err)
+	}
+	restore := replicationTestCapacity(obj)
+	t.Cleanup(func() {
+		cancel()
+		restore()
+		obj.Shutdown(context.Background())
+		removeRoots(dirs)
+	})
+	bucket := getRandomBucketName()
+	if err := obj.MakeBucket(ctx, bucket, MakeBucketOptions{VersioningEnabled: true}); err != nil {
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	er := z.serverPools[0].sets[0]
+	original := er.getDisks
+	t.Cleanup(func() { er.getDisks = original })
+	return z, er, append([]StorageAPI(nil), original()...), bucket
+}
+
+func seedPurgeMarker(t *testing.T, er *erasureObjects, bucket, name string, withData bool) (string, ObjectOptions) {
+	t.Helper()
+	dataVersion := ""
+	if withData {
+		oi, err := er.PutObject(t.Context(), bucket, name, mustGetPutObjReader(t, bytes.NewReader([]byte("data")), 4, "", ""), ObjectOptions{Versioned: true})
+		if err != nil {
+			t.Fatal(err)
+		}
+		dataVersion = oi.VersionID
+	}
+	opts := ObjectOptions{Versioned: true, VersionID: mustGetUUID(), DeleteMarker: true, ReplicationRequest: true, MTime: UTCNow().Add(-time.Hour), NoAuditLog: true}
+	opts.SetReplicaStatus(replication.Replica)
+	if _, err := er.DeleteObject(t.Context(), bucket, name, opts); err != nil {
+		t.Fatal(err)
+	}
+	opts.DeleteMarker = false
+	return dataVersion, opts
+}
+
+func countPurgeMarkers(t *testing.T, disks []StorageAPI, bucket, name, version string) int {
+	t.Helper()
+	n := 0
+	for _, disk := range disks {
+		fi, err := disk.ReadVersion(t.Context(), "", bucket, name, version, ReadOptions{})
+		if err == nil && fi.Deleted {
+			n++
+		} else if err != errFileNotFound && err != errFileVersionNotFound {
+			t.Fatalf("unexpected disk version: deleted=%v error=%v", fi.Deleted, err)
+		}
+	}
+	return n
+}
+
+func TestDeleteMarkerPurgeQuorum(t *testing.T) {
+	for _, tc := range []struct{ disks, online int }{{4, 2}, {16, 7}, {16, 8}, {16, 9}} {
+		t.Run(fmt.Sprintf("%d_disks_%d_online", tc.disks, tc.online), func(t *testing.T) {
+			z, er, all, bucket := markerPurgeFixture(t, tc.disks)
+			const name = "marker"
+			dataVersion, opts := seedPurgeMarker(t, er, bucket, name, true)
+			partial := make([]StorageAPI, len(all))
+			copy(partial, all[:tc.online])
+			er.getDisks = func() []StorageAPI { return partial }
+			_, first := z.DeleteObject(t.Context(), bucket, name, opts)
+			_, retry := z.DeleteObject(t.Context(), bucket, name, opts)
+			if tc.online <= tc.disks/2 {
+				if !isErrWriteQuorum(first) || !isErrWriteQuorum(retry) {
+					t.Errorf("below write quorum: first=%T %v, retry=%T %v", first, first, retry, retry)
+				}
+			} else if first != nil || (retry != nil && !isErrVersionNotFound(retry) && !isErrObjectNotFound(retry)) {
+				t.Errorf("write quorum available: first=%v retry=%v", first, retry)
+			}
+			want := tc.disks - tc.online
+			if tc.online < tc.disks/2 {
+				want = tc.disks
+			}
+			if got := countPurgeMarkers(t, all, bucket, name, opts.VersionID); got != want {
+				t.Errorf("partial purge markers=%d want=%d", got, want)
+			}
+			er.getDisks = func() []StorageAPI { return all }
+			_, err := z.DeleteObject(t.Context(), bucket, name, opts)
+			if err != nil && !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+				t.Errorf("full-online retry: %v", err)
+			}
+			if tc.online <= tc.disks/2 {
+				if got := countPurgeMarkers(t, all, bucket, name, opts.VersionID); got != 0 {
+					t.Errorf("full-online retry recreated/retained %d marker copies", got)
+				}
+			}
+			// A quorum-confirmed missing retry can leave minority residue. Exercise
+			// the real dangling-version healer after every disk has returned.
+			_, _ = er.HealObject(t.Context(), bucket, name, opts.VersionID, madmin.HealOpts{Remove: true})
+			if got := countPurgeMarkers(t, all, bucket, name, opts.VersionID); got != 0 {
+				t.Errorf("marker copies after retry and heal=%d", got)
+			}
+			oi, err := z.GetObjectInfo(t.Context(), bucket, name, ObjectOptions{Versioned: true})
+			if err != nil || oi.DeleteMarker || oi.VersionID != dataVersion {
+				t.Errorf("underlying version not visible: %+v error=%v", oi, err)
+			}
+		})
+	}
+}
+
+func TestDeleteMarkerPurgeMissingKey(t *testing.T) {
+	for _, pooled := range []bool{false, true} {
+		t.Run(fmt.Sprintf("pooled_%v", pooled), func(t *testing.T) {
+			z, er, all, bucket := markerPurgeFixture(t, 4)
+			_, opts := seedPurgeMarker(t, er, bucket, "marker-only", false)
+			er.getDisks = func() []StorageAPI { return []StorageAPI{all[0], all[1], nil, nil} }
+			deleteFn := er.DeleteObject
+			if pooled {
+				deleteFn = z.DeleteObject
+			}
+			_, _ = deleteFn(t.Context(), bucket, "marker-only", opts)
+			_, err := deleteFn(t.Context(), bucket, "marker-only", opts)
+			if !isErrWriteQuorum(err) {
+				t.Errorf("missing retry with only 2/4 absence votes returned %T %v", err, err)
+			}
+		})
+	}
+}
+
+func TestDeleteMarkerHealReplicationIdentity(t *testing.T) {
+	obj, er, all, bucket := markerPurgeFixture(t, 4)
+	const name = "healed-marker"
+	_, opts := seedPurgeMarker(t, er, bucket, name, true)
+	original, err := all[3].ReadVersion(t.Context(), "", bucket, name, opts.VersionID, ReadOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, disk := range all[:2] {
+		if err := disk.DeleteVersion(t.Context(), bucket, name, FileInfo{VersionID: opts.VersionID}, false, DeleteOptions{}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if _, err := er.HealObject(t.Context(), bucket, name, opts.VersionID, madmin.HealOpts{}); err != nil {
+		t.Fatal(err)
+	}
+	for i, disk := range all {
+		fi, err := disk.ReadVersion(t.Context(), "", bucket, name, opts.VersionID, ReadOptions{})
+		if err != nil || !maps.Equal(fi.Metadata, original.Metadata) {
+			t.Errorf("disk %d lost marker metadata: before=%v after=%v error=%v", i, original.Metadata, fi.Metadata, err)
+		}
+	}
+	// Only the healed half remains readable. Read actual disk state before
+	// invoking scanner replication, rather than constructing an ObjectInfo.
+	er.getDisks = func() []StorageAPI { return []StorageAPI{all[0], all[1], nil, nil} }
+	oi, err := er.GetObjectInfo(t.Context(), bucket, name, ObjectOptions{VersionID: opts.VersionID, Versioned: true})
+	if !oi.DeleteMarker || !isErrMethodNotAllowed(err) {
+		t.Fatalf("healed marker unreadable: %+v %v", oi, err)
+	}
+	er.getDisks = func() []StorageAPI { return all }
+	var outbound atomic.Int32
+	remote := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method == http.MethodDelete {
+			outbound.Add(1)
+			if r.Header.Get(xhttp.MinIOSourceDeleteMarker) != "true" || r.URL.Query().Get("versionId") != opts.VersionID {
+				t.Errorf("unexpected outbound request: %s %s", r.Method, r.URL)
+			}
+			w.WriteHeader(http.StatusNoContent)
+			return
+		}
+		w.WriteHeader(http.StatusNotFound)
+	}))
+	defer remote.Close()
+	client, err := minio.New(strings.TrimPrefix(remote.URL, "http://"), &minio.Options{Region: "us-east-1", MaxRetries: 1})
+	if err != nil {
+		t.Fatal(err)
+	}
+	arn := "arn:minio:replication::" + mustGetUUID() + ":bucket"
+	target := &TargetClient{Client: client, ARN: arn, Bucket: "target"}
+	oldTargets := globalBucketTargetSys
+	globalBucketTargetSys = &BucketTargetSys{arnRemotesMap: map[string]arnTarget{arn: {Client: target, lastRefresh: UTCNow()}}, targetsMap: map[string][]madmin.BucketTarget{bucket: {{Arn: arn, TargetBucket: "target"}}}, hc: map[string]epHealth{client.EndpointURL().Host: {Online: true}}}
+	defer func() { globalBucketTargetSys = oldTargets }()
+	cfg := configs[0]
+	cfg.RoleArn = arn
+	meta, err := globalBucketMetadataSys.Get(bucket)
+	if err != nil {
+		t.Fatal(err)
+	}
+	meta.replicationConfig = &cfg
+	globalBucketMetadataSys.Set(bucket, meta)
+	worker := make(chan ReplicationWorkerOperation, 1)
+	oldPool := globalReplicationPool
+	globalReplicationPool = once.NewSingleton[ReplicationPool]()
+	globalReplicationPool.Set(&ReplicationPool{ctx: t.Context(), objLayer: obj, workers: []chan ReplicationWorkerOperation{worker}, stats: globalReplicationStats.Load(), mrfSaveCh: make(chan MRFReplicateEntry, 1)})
+	defer func() { globalReplicationPool = oldPool }()
+	roi := queueReplicationHeal(t.Context(), bucket, oi, replicationConfig{Config: &cfg, remotes: &madmin.BucketTargets{Targets: []madmin.BucketTarget{{Arn: arn, TargetBucket: "target"}}}}, 0)
+	select {
+	case op := <-worker:
+		d := op.(DeletedObjectReplicationInfo)
+		result := replicateDeleteToTarget(t.Context(), d, target)
+		t.Errorf("healed replica scheduled for creation: version=%s existing=%v result=%+v", d.DeleteMarkerVersionID, roi.ExistingObjResync.mustResync(), result)
+	default:
+	}
+	if outbound.Load() != 0 {
+		t.Errorf("healed replica sent %d new marker creations", outbound.Load())
+	}
+}
@@ -0,0 +1,385 @@
+// Copyright (c) 2026 mr javad seydi and Ruohang Feng
+//
+// This file is part of Silo Object Storage stack.
+// SPDX-License-Identifier: AGPL-3.0-or-later
+
+package cmd
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"net/http"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/pgsty/silo-pkg/v3/policy"
+)
+
+const multipartScanEntryLimit = 100_000
+
+var (
+	multipartScanSlots          = make(chan struct{}, 2)
+	errMultipartListingLegacy   = errors.New("legacy multipart uploads require a coordinated upgrade and drain")
+	errMultipartListingIdentity = errors.New("multipart upload identity is invalid")
+)
+
+// One budget and admission slot cover the entire request, including all pools.
+// A slot is released only after the scan workers have actually stopped.
+type multipartScan struct {
+	ctx           context.Context
+	cancel        context.CancelFunc
+	remaining     atomic.Int64
+	metadataSlots chan struct{}
+	preflight     bool
+	sets          []multipartScanSet
+}
+
+type multipartScanSet struct {
+	Pool            int       `json:"pool"`
+	Set             int       `json:"set"`
+	Drives          int       `json:"drives"`
+	ScannedDrives   int       `json:"scannedDrives"`
+	UncoveredDrives []int     `json:"uncoveredDrives,omitempty"`
+	Candidates      int       `json:"candidates"`
+	LegacyUploads   int       `json:"legacyUploads"`
+	OldestLegacy    time.Time `json:"oldestLegacy,omitempty"`
+	Error           string    `json:"error,omitempty"`
+}
+
+type multipartPreflightReport struct {
+	Ready          bool               `json:"ready"`
+	Complete       bool               `json:"complete"`
+	Mode           string             `json:"mode"`
+	ScannedEntries int64              `json:"scannedEntries"`
+	LegacyUploads  int                `json:"legacyUploads"`
+	Sets           []multipartScanSet `json:"sets"`
+}
+
+func startMultipartScan(ctx context.Context, preflight bool) (*multipartScan, error) {
+	select {
+	case multipartScanSlots <- struct{}{}:
+	case <-ctx.Done():
+		return nil, ctx.Err()
+	default:
+		return nil, SlowDown{}
+	}
+	ctx, cancel := context.WithTimeout(ctx, 30*time.Second)
+	s := &multipartScan{ctx: ctx, cancel: cancel, preflight: preflight, metadataSlots: make(chan struct{}, multipartMetadataScanConcurrency)}
+	s.remaining.Store(multipartScanEntryLimit)
+	return s, nil
+}
+
+func (s *multipartScan) close() {
+	s.cancel()
+	<-multipartScanSlots
+}
+
+func (s *multipartScan) listDir(disk StorageAPI, bucket, dir string) ([]string, error) {
+	if err := s.ctx.Err(); err != nil {
+		return nil, SlowDown{}
+	}
+	remaining := s.remaining.Load()
+	if remaining <= 0 {
+		return nil, SlowDown{}
+	}
+	// Request one extra entry to detect overflow, never a silently partial page.
+	entries, err := disk.ListDir(s.ctx, bucket, minioMetaMultipartBucket, dir, int(remaining)+1)
+	if errors.Is(err, errFileNotFound) {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, err
+	}
+	if s.remaining.Add(-int64(len(entries))) < 0 {
+		return nil, SlowDown{}
+	}
+	return entries, nil
+}
+
+func (s *multipartScan) listUploadDirs(disk StorageAPI, bucket string) ([]string, error) {
+	hashDirs, err := s.listDir(disk, bucket, "")
+	if err != nil {
+		return nil, err
+	}
+	var candidates []string
+	for _, hashDir := range hashDirs {
+		if !strings.HasSuffix(hashDir, SlashSeparator) {
+			continue
+		}
+		hashDir = strings.TrimSuffix(hashDir, SlashSeparator)
+		uploadDirs, err := s.listDir(disk, bucket, hashDir)
+		if err != nil {
+			return nil, err
+		}
+		for _, uploadDir := range uploadDirs {
+			if strings.HasSuffix(uploadDir, SlashSeparator) {
+				candidates = append(candidates, pathJoin(hashDir, strings.TrimSuffix(uploadDir, SlashSeparator)))
+			}
+		}
+	}
+	return candidates, nil
+}
+
+// Identity is immutable at hash/uploadUUID. Check the hash before using even
+// a single source drive to exclude another bucket. Missing/bad identities must
+// fall back to the full metadata read; they cannot prove absence.
+func (er erasureObjects) multipartIdentity(fi FileInfo, shaDir string) (string, string, bool) {
+	bucket, object := fi.Metadata[multipartMetaBucket], fi.Metadata[multipartMetaObject]
+	return bucket, object, bucket != "" && object != "" && IsValidBucketName(bucket) &&
+		IsValidObjectPrefix(object) && er.getMultipartSHADir(bucket, object) == shaDir
+}
+
+func (er erasureObjects) readMultipartUploadCandidate(s *multipartScan, bucket, candidate string, source StorageAPI) (MultipartInfo, bool, bool, error) {
+	if s.ctx.Err() != nil {
+		return MultipartInfo{}, false, false, SlowDown{}
+	}
+	shaDir, uploadUUID, ok := strings.Cut(candidate, SlashSeparator)
+	if !ok || shaDir == "" || uploadUUID == "" || strings.Contains(uploadUUID, SlashSeparator) {
+		return MultipartInfo{}, false, false, errMultipartListingIdentity
+	}
+	if !s.preflight && bucket != "" && source != nil {
+		fi, err := source.ReadVersion(s.ctx, bucket, minioMetaMultipartBucket, candidate, "", ReadOptions{})
+		if err == nil {
+			storedBucket, _, valid := er.multipartIdentity(fi, shaDir)
+			if valid && storedBucket != bucket {
+				return MultipartInfo{}, false, false, nil
+			}
+		}
+	}
+	if s.ctx.Err() != nil {
+		return MultipartInfo{}, false, false, SlowDown{}
+	}
+	select {
+	case s.metadataSlots <- struct{}{}:
+		defer func() { <-s.metadataSlots }()
+	case <-s.ctx.Done():
+		return MultipartInfo{}, false, false, SlowDown{}
+	}
+	disks := er.getDisks()
+	metadata, errs := readAllFileInfo(s.ctx, disks, bucket, minioMetaMultipartBucket, candidate, "", false, false)
+	if s.preflight {
+		// Readiness is stronger than listing liveness: even one readable legacy
+		// copy must be drained, and an unreadable copy cannot certify readiness.
+		var oldest MultipartInfo
+		legacy := false
+		_, nativeID := multipartUploadTime(uploadUUID)
+		for i, err := range errs {
+			if errors.Is(err, errFileNotFound) || errors.Is(err, errFileVersionNotFound) {
+				continue
+			}
+			if err != nil {
+				return MultipartInfo{}, false, false, err
+			}
+			fi := metadata[i]
+			storedBucket, storedObject, valid := er.multipartIdentity(fi, shaDir)
+			if storedBucket != "" && storedObject != "" && !valid {
+				return MultipartInfo{}, false, false, errMultipartListingIdentity
+			}
+			if !valid || !nativeID {
+				info := multipartUploadInfo(storedBucket, storedObject, uploadUUID, fi.ModTime)
+				if !legacy || info.Initiated.Before(oldest.Initiated) {
+					oldest = info
+				}
+				legacy = true
+			}
+		}
+		if legacy {
+			return oldest, false, true, nil
+		}
+	}
+	readQuorum, _, err := objectQuorumFromMeta(s.ctx, metadata, errs, er.defaultParityCount)
+	if err != nil {
+		return MultipartInfo{}, false, false, err
+	}
+	_, modTime, etag := listOnlineDisks(disks, metadata, errs, readQuorum)
+	if err := reduceReadQuorumErrs(s.ctx, errs, objectOpIgnoredErrs, readQuorum); err != nil {
+		return MultipartInfo{}, false, false, err
+	}
+	fi, err := pickValidFileInfo(s.ctx, metadata, modTime, etag, readQuorum)
+	if err != nil {
+		return MultipartInfo{}, false, false, err
+	}
+	storedBucket, storedObject, valid := er.multipartIdentity(fi, shaDir)
+	info := multipartUploadInfo(storedBucket, storedObject, uploadUUID, fi.ModTime)
+	if storedBucket == "" || storedObject == "" {
+		return info, false, true, nil
+	}
+	if !valid {
+		return MultipartInfo{}, false, false, fmt.Errorf("%w: %s", errMultipartListingIdentity, candidate)
+	}
+	if _, ok := multipartUploadTime(uploadUUID); !ok {
+		return info, false, true, nil
+	}
+	if bucket != "" && bucket != storedBucket {
+		return MultipartInfo{}, false, false, nil
+	}
+	return info, true, false, nil
+}
+
+func (er erasureObjects) scanMultipartUploads(s *multipartScan, bucket string, poolIdx, setIdx int) ([]MultipartInfo, bool, error) {
+	disks := er.getDisks()
+	report := multipartScanSet{Pool: poolIdx, Set: setIdx, Drives: er.setDriveCount}
+	var resultErr error
+	defer func() {
+		if resultErr != nil {
+			report.Error = resultErr.Error()
+		}
+		s.sets = append(s.sets, report)
+	}()
+	var candidateMu sync.Mutex
+	candidates := make(map[string]StorageAPI)
+	errs := make([]error, len(disks))
+	var wg sync.WaitGroup
+	for i, disk := range disks {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			if disk == nil || !disk.IsOnline() {
+				errs[i] = errDiskNotFound
+				return
+			}
+			paths, err := s.listUploadDirs(disk, bucket)
+			errs[i] = err
+			if err != nil {
+				return
+			}
+			candidateMu.Lock()
+			defer candidateMu.Unlock()
+			for _, p := range paths {
+				candidates[p] = disk
+			}
+		}()
+	}
+	wg.Wait()
+	for i, err := range errs {
+		if err == nil {
+			report.ScannedDrives++
+		} else {
+			report.UncoveredDrives = append(report.UncoveredDrives, i)
+			if _, limited := err.(SlowDown); limited {
+				resultErr = err
+			}
+		}
+	}
+	report.Candidates = len(candidates)
+	// A successful scan must intersect every metadata read quorum, including
+	// records left with R copies by a partially failed cancellation.
+	if report.ScannedDrives < er.setDriveCount/2+1 && resultErr == nil {
+		resultErr = toObjectErr(errErasureReadQuorum, bucket)
+	}
+	if resultErr != nil {
+		return nil, false, resultErr
+	}
+	type candidate struct {
+		path   string
+		source StorageAPI
+	}
+	jobs := make(chan candidate)
+	var mu sync.Mutex
+	var uploads []MultipartInfo
+	for range min(16, len(candidates)) {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			for c := range jobs {
+				mu.Lock()
+				stopped := resultErr != nil
+				mu.Unlock()
+				if stopped || s.ctx.Err() != nil {
+					continue
+				}
+				upload, found, legacy, err := er.readMultipartUploadCandidate(s, bucket, c.path, c.source)
+				if errors.Is(err, errFileNotFound) || errors.Is(err, errFileVersionNotFound) {
+					continue
+				}
+				mu.Lock()
+				switch {
+				case err != nil && resultErr == nil:
+					resultErr = err
+				case legacy:
+					report.LegacyUploads++
+					if report.OldestLegacy.IsZero() || upload.Initiated.Before(report.OldestLegacy) {
+						report.OldestLegacy = upload.Initiated
+					}
+				case found && !s.preflight:
+					uploads = append(uploads, upload)
+				}
+				mu.Unlock()
+			}
+		}()
+	}
+dispatch:
+	for p, source := range candidates {
+		select {
+		case jobs <- candidate{p, source}:
+		case <-s.ctx.Done():
+			break dispatch
+		}
+	}
+	close(jobs)
+	wg.Wait()
+	if s.ctx.Err() != nil {
+		resultErr = SlowDown{}
+	}
+	return uploads, report.LegacyUploads != 0, resultErr
+}
+
+// multipartPreflight scans all pools, sets and drives, independent of caches.
+// A majority suffices for normal listing; upgrade readiness requires every
+// drive to have been inspected. The operator must also upgrade all writers.
+func (z *erasureServerPools) multipartPreflight(ctx context.Context) (multipartPreflightReport, error) {
+	s, err := startMultipartScan(ctx, true)
+	if err != nil {
+		return multipartPreflightReport{}, err
+	}
+	defer s.close()
+	report := multipartPreflightReport{Complete: true, Mode: "strict"}
+	if globalAPIConfig.getMultipartListingLegacy() {
+		report.Mode = "legacy"
+	}
+	for p, pool := range z.serverPools {
+		for i, set := range pool.sets {
+			_, _, _ = set.scanMultipartUploads(s, "", p, i)
+		}
+	}
+	report.Sets = s.sets
+	for _, set := range s.sets {
+		report.LegacyUploads += set.LegacyUploads
+		if set.Error != "" || set.ScannedDrives != set.Drives {
+			report.Complete = false
+		}
+	}
+	report.ScannedEntries = multipartScanEntryLimit - s.remaining.Load()
+	report.Ready = report.Complete && report.LegacyUploads == 0
+	return report, nil
+}
+
+// MultipartPreflightHandler is a read-only storage-admin diagnostic. It never
+// accepts an arbitrary deletion path or changes upload lifetime settings.
+func (a adminAPIHandlers) MultipartPreflightHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+	obj, _ := validateAdminReq(ctx, w, r, policy.StorageInfoAdminAction)
+	if obj == nil {
+		return
+	}
+	z, ok := obj.(*erasureServerPools)
+	if !ok {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrNotImplemented), r.URL)
+		return
+	}
+	report, err := z.multipartPreflight(ctx)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+	b, err := json.Marshal(report)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+	writeSuccessResponseJSON(w, b)
+}
@@ -0,0 +1,819 @@
+// Copyright (c) 2026 Ruohang Feng
+// SPDX-License-Identifier: AGPL-3.0-or-later
+
+package cmd
+
+import (
+	"context"
+	"encoding/base64"
+	"encoding/json"
+	"encoding/xml"
+	"errors"
+	"fmt"
+	"net/http"
+	"net/http/httptest"
+	"slices"
+	"strconv"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/minio/minio/internal/config/storageclass"
+)
+
+// Check the real handler and storage path, including a successful abort between
+// pages. Choose IDs increasing in both initiation time and lexical order so the
+// failure does not depend on differing interpretations of S3 marker ordering.
+func TestMultipartListingAbortBetweenHTTPPages(t *testing.T) {
+	z, _ := consistencyPools(t)
+	bucket, router, err := initAPIHandlerTest(t.Context(), z, []string{"ListMultipartUploads", "AbortMultipart"}, MakeBucketOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	setMultipartListingTestMode(t, false)
+	var firstID, secondID string
+	for attempt := 0; attempt < 32; attempt++ {
+		one, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+		if err != nil {
+			t.Fatal(err)
+		}
+		two, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+		if err != nil {
+			t.Fatal(err)
+		}
+		if one.UploadID < two.UploadID {
+			firstID, secondID = one.UploadID, two.UploadID
+			break
+		}
+		for _, id := range []string{one.UploadID, two.UploadID} {
+			if err := z.AbortMultipartUpload(t.Context(), bucket, "a", id, ObjectOptions{}); err != nil {
+				t.Fatal(err)
+			}
+		}
+	}
+	if firstID == "" {
+		t.Fatal("could not construct increasing upload IDs")
+	}
+	if _, err := z.NewMultipartUpload(t.Context(), bucket, "b", ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	request := func(method, u string) *httptest.ResponseRecorder {
+		t.Helper()
+		req, err := newTestSignedRequestV4(method, u, 0, nil, globalActiveCred.AccessKey, globalActiveCred.SecretKey, nil)
+		if err != nil {
+			t.Fatal(err)
+		}
+		rec := httptest.NewRecorder()
+		router.ServeHTTP(rec, req)
+		return rec
+	}
+	list := func(keyMarker, uploadMarker string, limit int) ListMultipartUploadsResponse {
+		t.Helper()
+		rec := request(http.MethodGet, getListMultipartUploadsURLWithParams("", bucket, "", keyMarker, uploadMarker, "", strconv.Itoa(limit)))
+		if rec.Code != http.StatusOK {
+			t.Fatalf("list HTTP %d: %s", rec.Code, rec.Body.String())
+		}
+		var result ListMultipartUploadsResponse
+		if err := xml.Unmarshal(rec.Body.Bytes(), &result); err != nil {
+			t.Fatal(err)
+		}
+		return result
+	}
+	first := list("", "", 1)
+	if len(first.Uploads) != 1 || first.Uploads[0].UploadID != firstID || !first.IsTruncated {
+		t.Fatalf("unexpected first page: %+v", first)
+	}
+	rec := request(http.MethodDelete, getAbortMultipartUploadURL("", bucket, "a", firstID))
+	if rec.Code != http.StatusNoContent {
+		t.Fatalf("abort HTTP %d: %s", rec.Code, rec.Body.String())
+	}
+	rest := list(first.NextKeyMarker, first.NextUploadIDMarker, 10)
+	var keys []string
+	for _, upload := range rest.Uploads {
+		keys = append(keys, upload.Key)
+	}
+	t.Logf("GET first page=200; DELETE marker=204; GET next page=200 keys=%v truncated=%v", keys, rest.IsTruncated)
+	if _, err := z.GetMultipartInfo(t.Context(), bucket, "a", secondID, ObjectOptions{}); err != nil {
+		t.Fatalf("remaining upload is not valid: %v", err)
+	}
+	if len(rest.Uploads) != 2 || rest.Uploads[0].UploadID != secondID {
+		t.Fatalf("valid remaining upload for key a is missing from continuation: %+v", rest.Uploads)
+	}
+}
+
+type multipartListingFaultDisk struct {
+	StorageAPI
+	read   func(context.Context, string, string, string, string, ReadOptions) (FileInfo, error)
+	delete func(context.Context, string, string, DeleteOptions) error
+}
+
+func (d multipartListingFaultDisk) ReadVersion(ctx context.Context, original, volume, object, version string, opts ReadOptions) (FileInfo, error) {
+	if d.read != nil {
+		return d.read(ctx, original, volume, object, version, opts)
+	}
+	return d.StorageAPI.ReadVersion(ctx, original, volume, object, version, opts)
+}
+
+func (d multipartListingFaultDisk) Delete(ctx context.Context, volume, object string, opts DeleteOptions) error {
+	if d.delete != nil {
+		return d.delete(ctx, volume, object, opts)
+	}
+	return d.StorageAPI.Delete(ctx, volume, object, opts)
+}
+
+func TestMultipartListingLegacyPreflight(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	const other = "multipart-legacy-other"
+	if err := z.MakeBucket(t.Context(), other, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	old, err := z.NewMultipartUpload(t.Context(), other, "old", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	fi, metadata, err := set.checkUploadIDExists(t.Context(), other, "old", old.UploadID, true)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for i := range metadata {
+		delete(metadata[i].Metadata, multipartMetaBucket)
+		delete(metadata[i].Metadata, multipartMetaObject)
+	}
+	if _, err = writeAllMetadata(t.Context(), set.getDisks(), other, minioMetaMultipartBucket,
+		set.getUploadIDDir(other, "old", old.UploadID), metadata, fi.WriteQuorum(set.defaultWQuorum())); err != nil {
+		t.Fatal(err)
+	}
+	if _, err = z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	z.mpCache.Clear()
+	t.Run("legacy-upgrade", func(t *testing.T) {
+		setMultipartListingTestMode(t, true)
+		exact, err := z.ListMultipartUploads(t.Context(), other, "old", "", "", "", 10)
+		if err != nil {
+			t.Fatal(err)
+		}
+		requireMultipartUploadKeys(t, exact, "old")
+		const empty = "multipart-legacy-empty"
+		if err := z.MakeBucket(t.Context(), empty, MakeBucketOptions{}); err != nil {
+			t.Fatal(err)
+		}
+		got, err := z.ListMultipartUploads(t.Context(), empty, "", "", "", "", 10)
+		if err != nil || len(got.Uploads) != 0 {
+			t.Fatalf("old upload in another bucket broke legacy listing: %+v %v", got, err)
+		}
+	})
+	_, err = z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 10)
+	if !errors.Is(err, errMultipartListingLegacy) {
+		t.Fatalf("old upload in another bucket: %v", err)
+	}
+	report, err := z.multipartPreflight(t.Context())
+	if err != nil || report.Ready || !report.Complete || report.LegacyUploads != 1 {
+		t.Fatalf("legacy preflight: %+v %v", report, err)
+	}
+	if err = z.AbortMultipartUpload(t.Context(), other, "old", old.UploadID, ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	report, err = z.multipartPreflight(t.Context())
+	if err != nil || !report.Ready || report.LegacyUploads != 0 {
+		t.Fatalf("drained preflight: %+v %v", report, err)
+	}
+	got, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, got, "a")
+}
+
+func TestMultipartListingIdentityFallback(t *testing.T) {
+	for _, kind := range []string{"old", "corrupt", "read-failure", "wrong-bucket"} {
+		t.Run(kind, func(t *testing.T) {
+			z, set, bucket := multipartListingFixture(t)
+			if _, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{}); err != nil {
+				t.Fatal(err)
+			}
+			original := set.getDisks
+			t.Cleanup(func() { set.getDisks = original })
+			var first atomic.Bool
+			set.getDisks = func() []StorageAPI {
+				disks := append([]StorageAPI(nil), original()...)
+				for i, d := range disks {
+					disks[i] = multipartListingFaultDisk{StorageAPI: d, read: func(ctx context.Context, b, v, p, version string, opts ReadOptions) (FileInfo, error) {
+						fi, err := d.ReadVersion(ctx, b, v, p, version, opts)
+						if v == minioMetaMultipartBucket && first.CompareAndSwap(false, true) {
+							if kind == "read-failure" {
+								return FileInfo{}, errDiskNotFound
+							}
+							if kind == "corrupt" {
+								return FileInfo{}, errFileCorrupt
+							}
+							fi.Metadata = cloneMSS(fi.Metadata)
+							if kind == "old" {
+								delete(fi.Metadata, multipartMetaBucket)
+							} else {
+								fi.Metadata[multipartMetaBucket] = "another-bucket"
+							}
+						}
+						return fi, err
+					}}
+				}
+				return disks
+			}
+			got, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 10)
+			if err != nil {
+				t.Fatal(err)
+			}
+			requireMultipartUploadKeys(t, got, "a")
+		})
+	}
+}
+
+func TestMultipartAbortPoolsAndRetry(t *testing.T) {
+	z, bucket := consistencyPools(t)
+	setMultipartListingTestMode(t, false)
+	mp, err := z.serverPools[1].NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	emptySet := z.serverPools[0].getHashedSet("a")
+	original := emptySet.getDisks
+	t.Cleanup(func() { emptySet.getDisks = original })
+	emptySet.getDisks = func() []StorageAPI {
+		disks := append([]StorageAPI(nil), original()...)
+		for i, d := range disks {
+			disks[i] = multipartListingFaultDisk{StorageAPI: d, read: func(context.Context, string, string, string, string, ReadOptions) (FileInfo, error) {
+				return FileInfo{}, errDiskNotFound
+			}}
+		}
+		return disks
+	}
+	err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{})
+	if err == nil || toAPIError(t.Context(), err).HTTPStatusCode != 503 {
+		t.Fatalf("unknown pool must not acknowledge cancellation: %v", err)
+	}
+	emptySet.getDisks = original
+	err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{})
+	if _, ok := err.(InvalidUploadID); !ok {
+		t.Fatalf("retry after all pools confirm absence: %v", err)
+	}
+	mp, err = z.serverPools[1].NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	if _, err = z.serverPools[1].GetMultipartInfo(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err == nil {
+		t.Fatal("empty first pool hid the actual upload")
+	}
+}
+
+func TestMultipartAbortRetryBelowReadQuorum(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	mp, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	original := set.getDisks
+	t.Cleanup(func() { set.getDisks = original })
+	set.getDisks = func() []StorageAPI {
+		disks := append([]StorageAPI(nil), original()...)
+		disks[3] = nil
+		d := disks[2]
+		disks[2] = multipartListingFaultDisk{StorageAPI: d, delete: func(ctx context.Context, v, p string, opts DeleteOptions) error {
+			if err := d.Delete(ctx, v, p, opts); err != nil {
+				return err
+			}
+			return context.DeadlineExceeded // operation finished, acknowledgement lost
+		}}
+		return disks
+	}
+	if err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err == nil {
+		t.Fatal("lost acknowledgement must fail")
+	}
+	set.getDisks = original
+	if err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+		t.Fatalf("one remaining metadata copy prevented retry: %v", err)
+	}
+}
+
+func TestMultipartListingMarkerHTTP(t *testing.T) {
+	z, _ := consistencyPools(t)
+	bucket, router, err := initAPIHandlerTest(t.Context(), z, []string{"ListMultipartUploads"}, MakeBucketOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	setMultipartListingTestMode(t, false)
+	for _, tc := range []struct {
+		key, marker string
+		status      int
+	}{
+		{"", "not-base64=", 200},
+		{"a", "not-base64=", 404},
+		{"a", base64.RawURLEncoding.EncodeToString([]byte("not-native")), 400},
+		{"a", multipartListingTestID(time.Unix(100, 0), 1), 200},
+	} {
+		u := getListMultipartUploadsURLWithParams("", bucket, "", tc.key, tc.marker, "", "10")
+		req, err := newTestSignedRequestV4(http.MethodGet, u, 0, nil, globalActiveCred.AccessKey, globalActiveCred.SecretKey, nil)
+		if err != nil {
+			t.Fatal(err)
+		}
+		rec := httptest.NewRecorder()
+		router.ServeHTTP(rec, req)
+		if rec.Code != tc.status {
+			t.Fatalf("key=%q marker=%q: %d %s", tc.key, tc.marker, rec.Code, rec.Body.String())
+		}
+	}
+}
+
+func TestMultipartPreflightAdminHTTP(t *testing.T) {
+	bed, err := prepareAdminErasureTestBed(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { bed.done(); bed.objLayer.Shutdown(context.Background()); removeRoots(bed.erasureDirs) })
+	const target = "/minio/admin/v3/multipart-preflight"
+	rec := httptest.NewRecorder()
+	bed.router.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, target, nil))
+	if rec.Code != 403 {
+		t.Fatalf("anonymous preflight: %d", rec.Code)
+	}
+	req, err := newTestSignedRequestV4(http.MethodGet, target, 0, nil, globalActiveCred.AccessKey, globalActiveCred.SecretKey, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	rec = httptest.NewRecorder()
+	bed.router.ServeHTTP(rec, req)
+	if rec.Code != 200 {
+		t.Fatalf("admin preflight: %d %s", rec.Code, rec.Body.String())
+	}
+	var report multipartPreflightReport
+	if err = json.Unmarshal(rec.Body.Bytes(), &report); err != nil || !report.Ready || !report.Complete {
+		t.Fatalf("preflight: %+v %v", report, err)
+	}
+	for range cap(multipartScanSlots) {
+		scan, err := startMultipartScan(t.Context(), false)
+		if err != nil {
+			t.Fatal(err)
+		}
+		defer scan.close()
+	}
+	rec = httptest.NewRecorder()
+	bed.router.ServeHTTP(rec, req)
+	if rec.Code != http.StatusServiceUnavailable {
+		t.Fatalf("busy admin preflight: %d %s", rec.Code, rec.Body.String())
+	}
+}
+
+type multipartLateCreateDisk struct {
+	StorageAPI
+	late chan<- func() error
+}
+
+func (d multipartLateCreateDisk) WriteMetadata(_ context.Context, original, volume, object string, fi FileInfo) error {
+	if volume == minioMetaMultipartBucket {
+		d.late <- func() error { return d.StorageAPI.WriteMetadata(context.Background(), original, volume, object, fi) }
+		return context.DeadlineExceeded
+	}
+	return d.StorageAPI.WriteMetadata(context.Background(), original, volume, object, fi)
+}
+
+// This characterization is deliberately NOT an assertion of terminal abort
+// correctness. It preserves an executable example of the separately scoped
+// late-creation-write limitation. Change the expectation when fencing is added.
+func TestMultipartAbortLateCreateBoundary(t *testing.T) {
+	obj, dirs, err := prepareErasure(t.Context(), 16)
+	if err != nil {
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	setMultipartListingTestMode(t, false)
+	t.Cleanup(func() { z.Shutdown(context.Background()); removeRoots(dirs) })
+	saved := globalStorageClass
+	globalStorageClass.Update(storageclass.Config{Standard: storageclass.StorageClass{Parity: 8}})
+	t.Cleanup(func() { globalStorageClass.Update(saved) })
+	const bucket = "multipart-late-create"
+	if err = z.MakeBucket(t.Context(), bucket, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	set := z.serverPools[0].getHashedSet("a")
+	original := set.getDisks
+	t.Cleanup(func() { set.getDisks = original })
+	late := make(chan func() error, 16)
+	set.getDisks = func() []StorageAPI {
+		disks := append([]StorageAPI(nil), original()...)
+		for i := 9; i < 16; i++ {
+			disks[i] = multipartLateCreateDisk{disks[i], late}
+		}
+		return disks
+	}
+	mp, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(late) != 7 {
+		t.Fatalf("expected seven timed-out creation writes, got %d", len(late))
+	}
+	set.getDisks = func() []StorageAPI {
+		disks := append([]StorageAPI(nil), original()...)
+		for i := 2; i < 9; i++ {
+			disks[i] = nil
+		}
+		return disks
+	}
+	if err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	for range 7 {
+		if err = (<-late)(); err != nil {
+			t.Fatal(err)
+		}
+	}
+	set.getDisks = original
+	_, metadata, err := set.checkUploadIDExists(t.Context(), bucket, "a", mp.UploadID, true)
+	if err != nil {
+		t.Fatalf("late creation boundary changed; review and remove the documented limitation: %v", err)
+	}
+	count := 0
+	for _, fi := range metadata {
+		if fi.IsValid() {
+			count++
+		}
+	}
+	if count != 14 {
+		t.Fatalf("expected fourteen resurrected copies, got %d", count)
+	}
+	t.Log("KNOWN UNRESOLVED BOUNDARY: seven delayed creation writes plus seven offline copies restore a writable upload after acknowledged cancellation")
+}
+
+type multipartListingCountingDisk struct {
+	StorageAPI
+	directoryCalls *atomic.Int64
+	metadataCalls  *atomic.Int64
+}
+
+func (d multipartListingCountingDisk) ListDir(ctx context.Context, original, volume, dir string, count int) ([]string, error) {
+	if volume == minioMetaMultipartBucket {
+		d.directoryCalls.Add(1)
+	}
+	return d.StorageAPI.ListDir(ctx, original, volume, dir, count)
+}
+
+func (d multipartListingCountingDisk) ReadVersion(ctx context.Context, original, volume, object, version string, opts ReadOptions) (FileInfo, error) {
+	if volume == minioMetaMultipartBucket {
+		d.metadataCalls.Add(1)
+	}
+	return d.StorageAPI.ReadVersion(ctx, original, volume, object, version, opts)
+}
+
+// Counts storage API calls; this is not a deployment throughput benchmark.
+func TestMultipartListingScanCosts(t *testing.T) {
+	obj, dirs, err := prepareErasure(t.Context(), 4)
+	if err != nil {
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	setMultipartListingTestMode(t, false)
+	t.Cleanup(func() { z.Shutdown(context.Background()); removeRoots(dirs) })
+	const bucket, otherBucket = "r9-scan-target", "r9-scan-unrelated"
+	for _, name := range []string{bucket, otherBucket} {
+		if err := z.MakeBucket(t.Context(), name, MakeBucketOptions{}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if _, err := z.NewMultipartUpload(t.Context(), bucket, "dir/target", ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	var directoryCalls, metadataCalls atomic.Int64
+	for _, pool := range z.serverPools {
+		for _, set := range pool.sets {
+			original := set.getDisks
+			set.getDisks = func() []StorageAPI {
+				disks := original()
+				wrapped := make([]StorageAPI, len(disks))
+				for i, disk := range disks {
+					if disk != nil {
+						wrapped[i] = multipartListingCountingDisk{disk, &directoryCalls, &metadataCalls}
+					}
+				}
+				return wrapped
+			}
+			t.Cleanup(func() { set.getDisks = original })
+		}
+	}
+	measure := func(label string) (int64, int64) {
+		t.Helper()
+		directoryCalls.Store(0)
+		metadataCalls.Store(0)
+		result, err := z.ListMultipartUploads(t.Context(), bucket, "dir/", "", "", "", 1)
+		if err != nil {
+			t.Fatal(err)
+		}
+		requireMultipartUploadKeys(t, result, "dir/target")
+		d, m := directoryCalls.Load(), metadataCalls.Load()
+		t.Logf("%s: max-uploads=1 returned=%d ListDir=%d ReadVersion=%d", label, len(result.Uploads), d, m)
+		return d, m
+	}
+	_, baseline := measure("no unrelated uploads")
+	for n := 0; n < 32; n++ {
+		if _, err := z.NewMultipartUpload(t.Context(), otherBucket, fmt.Sprintf("other/%03d", n), ObjectOptions{}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	_, loaded := measure("32 uploads in another bucket")
+	_, repeated := measure("same query repeated")
+	if loaded != baseline+32 || repeated != loaded {
+		t.Fatalf("unexpected scan accounting: baseline=%d loaded=%d repeated=%d", baseline, loaded, repeated)
+	}
+}
+
+func multipartListingTestID(created time.Time, n int) string {
+	return base64.RawURLEncoding.EncodeToString([]byte(fmt.Sprintf("00000000-0000-4000-8000-000000000000.00000000-0000-4000-8000-%012xx%d", n, created.UnixNano())))
+}
+
+func multipartListingFixture(t *testing.T) (*erasureServerPools, *erasureObjects, string) {
+	t.Helper()
+	obj, dirs, err := prepareErasure(t.Context(), 4)
+	if err != nil {
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	t.Cleanup(func() { z.Shutdown(context.Background()); removeRoots(dirs) })
+	const bucket = "multipart-listing-test"
+	if err := z.MakeBucket(t.Context(), bucket, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	setMultipartListingTestMode(t, false)
+	return z, z.serverPools[0].getHashedSet("a"), bucket
+}
+
+func TestMultipartListingMissingMarkers(t *testing.T) {
+	base := time.Unix(100, 0)
+	sameTime := []string{multipartListingTestID(base.Add(time.Second), 1), multipartListingTestID(base.Add(time.Second), 2), multipartListingTestID(base.Add(time.Second), 3)}
+	slices.Sort(sameTime)
+	uploads := []MultipartInfo{
+		{Bucket: "bucket", Object: "a", UploadID: multipartListingTestID(base, 1), Initiated: base},
+		{Bucket: "bucket", Object: "a", UploadID: sameTime[1], Initiated: base.Add(time.Second)},
+		{Bucket: "bucket", Object: "b", UploadID: multipartListingTestID(base, 4), Initiated: base},
+	}
+	for _, tc := range []struct {
+		name    string
+		created time.Time
+		id      int
+		want    []string
+	}{
+		{"before", base.Add(-time.Second), 0, []string{"a", "a", "b"}},
+		{"between", base.Add(time.Second / 2), 2, []string{"a", "b"}},
+		{"same-time-before", base.Add(time.Second), 2, []string{"a", "b"}},
+		{"same-time-after", base.Add(time.Second), 4, []string{"b"}},
+		{"after", base.Add(2 * time.Second), 5, []string{"b"}},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			marker := multipartListingTestID(tc.created, tc.id)
+			if tc.name == "same-time-before" {
+				marker = sameTime[0]
+			}
+			if tc.name == "same-time-after" {
+				marker = sameTime[2]
+			}
+			if err := checkListMultipartArgs(t.Context(), "bucket", "", "a", marker, ""); err != nil {
+				t.Fatal(err)
+			}
+			got := paginateMultipartUploads(uploads, "", "a", marker, "", 10)
+			if !slices.Equal(multipartUploadKeys(got.Uploads), tc.want) {
+				t.Fatalf("got %v want %v", multipartUploadKeys(got.Uploads), tc.want)
+			}
+		})
+	}
+}
+
+func TestMultipartListingAbortRecovery(t *testing.T) {
+	for _, offline := range []int{1, 2} {
+		t.Run(fmt.Sprint(offline), func(t *testing.T) {
+			z, set, bucket := multipartListingFixture(t)
+			mp, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			original := set.getDisks
+			t.Cleanup(func() { set.getDisks = original })
+			set.getDisks = func() []StorageAPI {
+				disks := append([]StorageAPI(nil), original()...)
+				for i := 0; i < offline; i++ {
+					disks[i] = nil
+				}
+				return disks
+			}
+			err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{})
+			if offline == 1 && err != nil {
+				t.Fatal(err)
+			}
+			if offline == 2 && (err == nil || toAPIError(t.Context(), err).HTTPStatusCode != 503) {
+				t.Fatalf("two offline drives must not acknowledge cancellation: %v", err)
+			}
+			set.getDisks = original
+			if offline == 2 {
+				// Retry a partially completed deletion after the original disks return.
+				if err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatal(err)
+				}
+			}
+			got, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 100)
+			if err != nil || len(got.Uploads) != 0 {
+				t.Fatalf("canceled upload reappeared: %+v %v", got, err)
+			}
+		})
+	}
+}
+
+func TestMultipartListingCoverage(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	mp, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	original := set.getDisks
+	t.Cleanup(func() { set.getDisks = original })
+	disks := original()
+	// Leave a readable two-copy record, simulating a partially failed abort.
+	for _, d := range disks[:2] {
+		if err = d.Delete(t.Context(), minioMetaMultipartBucket, set.getUploadIDDir(bucket, "a", mp.UploadID), DeleteOptions{Recursive: true}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	set.getDisks = func() []StorageAPI { return []StorageAPI{disks[0], disks[1], nil, nil} }
+	_, err = z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 10)
+	if err == nil || toAPIError(t.Context(), err).HTTPStatusCode != 503 {
+		t.Fatalf("incomplete discovery returned success: %v", err)
+	}
+	set.getDisks = func() []StorageAPI { return []StorageAPI{nil, disks[1], disks[2], disks[3]} }
+	got, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, got, "a")
+	report, err := z.multipartPreflight(t.Context())
+	if err != nil || report.Ready || report.Complete || len(report.Sets[0].UncoveredDrives) != 1 {
+		t.Fatalf("offline upgrade preflight: %+v %v", report, err)
+	}
+}
+
+func TestMultipartListingBudgetAndAdmission(t *testing.T) {
+	_, set, bucket := multipartListingFixture(t)
+	if _, err := set.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	scan, err := startMultipartScan(t.Context(), false)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer scan.close()
+	scan.remaining.Store(1)
+	_, _, err = set.scanMultipartUploads(scan, bucket, 0, 0)
+	var limited SlowDown
+	if !errors.As(err, &limited) {
+		t.Fatalf("budget: %v", err)
+	}
+	if apiErr := toAPIError(t.Context(), err); apiErr.HTTPStatusCode != http.StatusServiceUnavailable || apiErr.Code != "SlowDown" {
+		t.Fatalf("budget error mapping: %+v", apiErr)
+	}
+	second, err := startMultipartScan(t.Context(), false)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer second.close()
+	if third, err := startMultipartScan(t.Context(), false); err == nil {
+		third.close()
+		t.Fatal("third scan admitted")
+	}
+}
+
+func TestMultipartListingAdmissionHTTP(t *testing.T) {
+	z, _, _ := multipartListingFixture(t)
+	bucket, router, err := initAPIHandlerTest(t.Context(), z, []string{"ListMultipartUploads"}, MakeBucketOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	setMultipartListingTestMode(t, false)
+	for range cap(multipartScanSlots) {
+		scan, err := startMultipartScan(t.Context(), false)
+		if err != nil {
+			t.Fatal(err)
+		}
+		defer scan.close()
+	}
+	req, err := newTestSignedRequestV4(http.MethodGet,
+		getListMultipartUploadsURLWithParams("", bucket, "", "", "", "", "1"),
+		0, nil, globalActiveCred.AccessKey, globalActiveCred.SecretKey, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	rec := httptest.NewRecorder()
+	router.ServeHTTP(rec, req)
+	var response APIErrorResponse
+	if err := xml.Unmarshal(rec.Body.Bytes(), &response); err != nil {
+		t.Fatal(err)
+	}
+	if rec.Code != http.StatusServiceUnavailable || response.Code != "SlowDown" {
+		t.Fatalf("admission returned %d %s", rec.Code, rec.Body.String())
+	}
+}
+
+func TestMultipartListingPreflightMinorityLegacy(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	mp, err := z.NewMultipartUpload(t.Context(), bucket, "old", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	path := set.getUploadIDDir(bucket, "old", mp.UploadID)
+	disks := set.getDisks()
+	fi, err := disks[0].ReadVersion(t.Context(), bucket, minioMetaMultipartBucket, path, "", ReadOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	delete(fi.Metadata, multipartMetaBucket)
+	delete(fi.Metadata, multipartMetaObject)
+	if err = disks[0].WriteMetadata(t.Context(), bucket, minioMetaMultipartBucket, path, fi); err != nil {
+		t.Fatal(err)
+	}
+	for _, d := range disks[1:] {
+		if err = d.Delete(t.Context(), minioMetaMultipartBucket, path, DeleteOptions{Recursive: true}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	report, err := z.multipartPreflight(t.Context())
+	if err != nil || report.Ready || !report.Complete || report.LegacyUploads != 1 {
+		t.Fatalf("minority legacy copy must prevent readiness: %+v %v", report, err)
+	}
+}
+
+func TestMultipartListingCancellationRetainsAdmission(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	for i := range 40 {
+		if _, err := z.NewMultipartUpload(t.Context(), bucket, fmt.Sprintf("key-%02d", i), ObjectOptions{}); err != nil {
+			t.Fatal(err)
+		}
+	}
+	original := set.getDisks
+	t.Cleanup(func() { set.getDisks = original })
+	entered := make(chan struct{}, 40)
+	release := make(chan struct{})
+	var once sync.Once
+	defer once.Do(func() { close(release) })
+	var reads atomic.Int32
+	set.getDisks = func() []StorageAPI {
+		disks := append([]StorageAPI(nil), original()...)
+		for i, d := range disks {
+			disks[i] = multipartListingFaultDisk{StorageAPI: d, read: func(ctx context.Context, b, v, p, version string, opts ReadOptions) (FileInfo, error) {
+				if v != minioMetaMultipartBucket {
+					return d.ReadVersion(ctx, b, v, p, version, opts)
+				}
+				reads.Add(1)
+				entered <- struct{}{}
+				// Model an RPC which does not return immediately on cancellation.
+				<-release
+				return FileInfo{}, ctx.Err()
+			}}
+		}
+		return disks
+	}
+	second, err := startMultipartScan(t.Context(), false)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer second.close()
+	ctx, cancel := context.WithCancel(t.Context())
+	defer cancel()
+	done := make(chan error, 1)
+	go func() {
+		_, err := z.ListMultipartUploads(ctx, bucket, "", "", "", "", 10)
+		done <- err
+	}()
+	for range 16 {
+		select {
+		case <-entered:
+		case <-time.After(5 * time.Second):
+			t.Fatal("identity workers did not start")
+		}
+	}
+	cancel()
+	if extra, err := startMultipartScan(t.Context(), false); err == nil {
+		extra.close()
+		t.Fatal("canceled request released admission while RPCs were still running")
+	}
+	start := time.Now()
+	once.Do(func() { close(release) })
+	select {
+	case err := <-done:
+		if err == nil {
+			t.Fatal("canceled scan returned a successful partial list")
+		}
+	case <-time.After(2 * time.Second):
+		t.Fatal("scan did not stop after blocked RPCs returned")
+	}
+	if got := reads.Load(); got != 16 {
+		t.Fatalf("scheduled more identity/metadata reads after cancellation: %d", got)
+	}
+	t.Logf("16 identity RPCs bounded; admission held until return; cancellation settled in %s", time.Since(start))
+}
@@ -0,0 +1,331 @@
+// Copyright (c) 2026 Ruohang Feng
+// SPDX-License-Identifier: AGPL-3.0-or-later
+
+package cmd
+
+import (
+	"context"
+	"encoding/base64"
+	"errors"
+	"fmt"
+	"strings"
+	"sync/atomic"
+	"testing"
+
+	"github.com/minio/minio/internal/config/api"
+	"github.com/minio/minio/internal/config/storageclass"
+	"github.com/minio/minio/internal/dsync"
+	"github.com/minio/minio/internal/grid"
+	xnet "github.com/pgsty/silo-pkg/v3/net"
+)
+
+func setMultipartListingTestMode(t *testing.T, legacy bool) {
+	t.Helper()
+	globalAPIConfig.mu.Lock()
+	previous := globalAPIConfig.multipartListingStrict
+	globalAPIConfig.multipartListingStrict = !legacy
+	globalAPIConfig.mu.Unlock()
+	t.Cleanup(func() {
+		globalAPIConfig.mu.Lock()
+		globalAPIConfig.multipartListingStrict = previous
+		globalAPIConfig.mu.Unlock()
+	})
+}
+
+func TestMultipartListingDefaultMode(t *testing.T) {
+	var uninitialized apiConfig
+	if !uninitialized.getMultipartListingLegacy() {
+		t.Fatal("runtime zero value enabled strict mode before config initialization")
+	}
+	for _, mode := range []string{"", "legacy", "strict"} {
+		var local apiConfig
+		local.init(api.Config{RequestsMax: 1, MultipartListing: mode}, []int{4}, false)
+		if local.getMultipartListingLegacy() != (mode != "strict") {
+			t.Fatalf("unexpected mode for %q", mode)
+		}
+	}
+}
+
+func TestMultipartAbortLegacyAvailability(t *testing.T) {
+	for _, drives := range []int{4, 16} {
+		for _, offline := range []int{0, 1, drives / 2} {
+			t.Run(fmt.Sprintf("drives=%d/offline=%d", drives, offline), func(t *testing.T) {
+				obj, dirs, err := prepareErasure(t.Context(), drives)
+				if err != nil {
+					t.Fatal(err)
+				}
+				z := obj.(*erasureServerPools)
+				t.Cleanup(func() { z.Shutdown(context.Background()); removeRoots(dirs) })
+				setMultipartListingTestMode(t, true)
+				previous := globalStorageClass
+				globalStorageClass.Update(storageclass.Config{Standard: storageclass.StorageClass{Parity: drives / 2}})
+				t.Cleanup(func() { globalStorageClass.Update(previous) })
+				const bucket, key = "multipart-legacy-availability", "keep-available"
+				if err := z.MakeBucket(t.Context(), bucket, MakeBucketOptions{}); err != nil {
+					t.Fatal(err)
+				}
+				mp, err := z.NewMultipartUpload(t.Context(), bucket, key, ObjectOptions{})
+				if err != nil {
+					t.Fatal(err)
+				}
+				set := z.serverPools[0].getHashedSet(key)
+				original := set.getDisks
+				disks := original()
+				set.getDisks = func() []StorageAPI {
+					visible := append([]StorageAPI(nil), disks...)
+					clear(visible[:offline])
+					return visible
+				}
+				t.Cleanup(func() { set.getDisks = original })
+				if err := z.AbortMultipartUpload(t.Context(), bucket, key, mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatalf("released read-quorum availability regressed: %v", err)
+				}
+				for _, disk := range disks[offline:] {
+					_, err := disk.ReadVersion(t.Context(), bucket, minioMetaMultipartBucket, set.getUploadIDDir(bucket, key, mp.UploadID), "", ReadOptions{})
+					if !errors.Is(err, errFileNotFound) {
+						t.Fatalf("online replica not cleaned: %v", err)
+					}
+				}
+			})
+		}
+	}
+}
+
+func TestMultipartAbortLegacyPoolOrder(t *testing.T) {
+	for _, owner := range []int{0, 1} {
+		t.Run(fmt.Sprintf("owner=%d", owner), func(t *testing.T) {
+			z, bucket := consistencyPools(t)
+			setMultipartListingTestMode(t, true)
+			mp, err := z.serverPools[owner].NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			emptySet := z.serverPools[1-owner].getHashedSet("a")
+			original := emptySet.getDisks
+			t.Cleanup(func() { emptySet.getDisks = original })
+			emptySet.getDisks = func() []StorageAPI { return make([]StorageAPI, emptySet.setDriveCount) }
+			err = z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{})
+			if owner == 0 && err != nil {
+				t.Fatalf("unrelated later pool blocked legacy cancellation: %v", err)
+			}
+			if owner == 1 {
+				if err == nil || toAPIError(t.Context(), err).HTTPStatusCode != 503 {
+					t.Fatalf("unknown earlier pool must retain released error behavior: %v", err)
+				}
+				if _, err := z.serverPools[owner].GetMultipartInfo(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatalf("later pool visited despite earlier error: %v", err)
+				}
+				emptySet.getDisks = original
+				if err := z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatal(err)
+				}
+			}
+		})
+	}
+}
+
+func TestMultipartAbortMinorityLegacyCleanup(t *testing.T) {
+	for _, failDelete := range []bool{false, true} {
+		t.Run(fmt.Sprintf("delete-fails=%v", failDelete), func(t *testing.T) {
+			z, set, bucket := multipartListingFixture(t)
+			mp, err := z.NewMultipartUpload(t.Context(), bucket, "old", ObjectOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			path := set.getUploadIDDir(bucket, "old", mp.UploadID)
+			original := set.getDisks
+			disks := original()
+			fi, err := disks[0].ReadVersion(t.Context(), bucket, minioMetaMultipartBucket, path, "", ReadOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			delete(fi.Metadata, multipartMetaBucket)
+			delete(fi.Metadata, multipartMetaObject)
+			if err := disks[0].WriteMetadata(t.Context(), bucket, minioMetaMultipartBucket, path, fi); err != nil {
+				t.Fatal(err)
+			}
+			for _, d := range disks[1:] {
+				if err := d.Delete(t.Context(), minioMetaMultipartBucket, path, DeleteOptions{Recursive: true}); err != nil {
+					t.Fatal(err)
+				}
+			}
+			report, err := z.multipartPreflight(t.Context())
+			if err != nil || report.Ready || report.LegacyUploads != 1 {
+				t.Fatalf("invalid legacy fixture: %+v %v", report, err)
+			}
+			if failDelete {
+				set.getDisks = func() []StorageAPI {
+					wrapped := append([]StorageAPI(nil), disks...)
+					wrapped[0] = multipartListingFaultDisk{StorageAPI: disks[0], delete: func(context.Context, string, string, DeleteOptions) error { return errFaultyDisk }}
+					return wrapped
+				}
+				t.Cleanup(func() { set.getDisks = original })
+				err := z.AbortMultipartUpload(t.Context(), bucket, "old", mp.UploadID, ObjectOptions{})
+				if err == nil || toAPIError(t.Context(), err).HTTPStatusCode != 503 {
+					t.Fatalf("empty drives masked failed deletion of the observed replica: %v", err)
+				}
+				if _, present := z.mpCache.Load(mp.UploadID); !present {
+					t.Fatal("failed cancellation evicted upload from cache")
+				}
+				set.getDisks = original
+			}
+			if err := z.AbortMultipartUpload(t.Context(), bucket, "old", mp.UploadID, ObjectOptions{}); err != nil {
+				t.Fatal(err)
+			}
+			report, err = z.multipartPreflight(t.Context())
+			if err != nil || !report.Ready || report.LegacyUploads != 0 {
+				t.Fatalf("acknowledged cleanup left online legacy remnants: %+v %v", report, err)
+			}
+			if _, present := z.mpCache.Load(mp.UploadID); present {
+				t.Fatal("successful cancellation retained cache entry")
+			}
+		})
+	}
+}
+
+func TestMultipartAbortWrongTargetKeepsCache(t *testing.T) {
+	for _, legacy := range []bool{true, false} {
+		t.Run(fmt.Sprintf("legacy=%v", legacy), func(t *testing.T) {
+			z, _, bucket := multipartListingFixture(t)
+			setMultipartListingTestMode(t, legacy)
+			const other = "multipart-other-bucket"
+			if err := z.MakeBucket(t.Context(), other, MakeBucketOptions{}); err != nil {
+				t.Fatal(err)
+			}
+			mp, err := z.NewMultipartUpload(t.Context(), bucket, "valid-key", ObjectOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			for _, target := range [][2]string{{bucket, "wrong-key"}, {other, "valid-key"}} {
+				err := z.AbortMultipartUpload(t.Context(), target[0], target[1], mp.UploadID, ObjectOptions{})
+				var invalid InvalidUploadID
+				if !errors.As(err, &invalid) {
+					t.Fatalf("wrong target result: %v", err)
+				}
+				if _, present := z.mpCache.Load(mp.UploadID); !present {
+					t.Fatal("wrong-target abort evicted another upload")
+				}
+				if _, err := z.GetMultipartInfo(t.Context(), bucket, "valid-key", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatalf("wrong-target abort harmed valid upload: %v", err)
+				}
+			}
+			if err := z.AbortMultipartUpload(t.Context(), bucket, "valid-key", mp.UploadID, ObjectOptions{}); err != nil {
+				t.Fatal(err)
+			}
+			var invalid InvalidUploadID
+			if err := z.AbortMultipartUpload(t.Context(), bucket, "valid-key", mp.UploadID, ObjectOptions{}); !errors.As(err, &invalid) {
+				t.Fatalf("repeat abort: %v", err)
+			}
+		})
+	}
+}
+
+func TestMultipartAbortRejectsUnsafeID(t *testing.T) {
+	for _, legacy := range []bool{true, false} {
+		t.Run(fmt.Sprintf("legacy=%v", legacy), func(t *testing.T) {
+			z, _, bucket := multipartListingFixture(t)
+			setMultipartListingTestMode(t, legacy)
+			mp, err := z.NewMultipartUpload(t.Context(), bucket, "keep", ObjectOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			for _, suffix := range []string{".", "..", "../other", "a/b", "a\\b", ""} {
+				id := base64.RawURLEncoding.EncodeToString([]byte("deployment." + suffix))
+				err := z.AbortMultipartUpload(t.Context(), bucket, "keep", id, ObjectOptions{})
+				var invalid InvalidUploadID
+				if !errors.As(err, &invalid) {
+					t.Fatalf("unsafe ID accepted: suffix=%q err=%v", suffix, err)
+				}
+				if _, err := z.GetMultipartInfo(t.Context(), bucket, "keep", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatalf("valid upload harmed by rejected ID: %v", err)
+				}
+			}
+		})
+	}
+}
+
+func TestMultipartAbortPeerNotification(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	tg, err := grid.SetupTestGrid(2)
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(tg.Cleanup)
+	var notifications atomic.Int32
+	if err := cleanupUploadIDCacheMetaRPC.Register(tg.Managers[1], func(*grid.MSS) (grid.NoPayload, *grid.RemoteErr) {
+		notifications.Add(1)
+		return grid.NoPayload{}, nil
+	}); err != nil {
+		t.Fatal(err)
+	}
+	host, err := xnet.ParseHost(strings.TrimPrefix(tg.Hosts[1], "http://"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	previous := globalNotificationSys
+	globalNotificationSys = &NotificationSys{peerClients: []*peerRESTClient{{
+		host:     host,
+		gridConn: func() *grid.Connection { return tg.Managers[0].Connection(tg.Hosts[1]) },
+	}}}
+	t.Cleanup(func() { globalNotificationSys = previous })
+	// Use the real distributed lock implementation: Unlock cancels its derived
+	// context. Local locks do not, and would hide a broken notification context.
+	oldMutex, oldLockers := set.nsMutex, set.getLockers
+	lockers := []dsync.NetLocker{newLocker(), newLocker()}
+	set.nsMutex = newNSLock(true)
+	set.getLockers = func() ([]dsync.NetLocker, string) { return lockers, "multipart-test" }
+	t.Cleanup(func() { set.nsMutex, set.getLockers = oldMutex, oldLockers })
+	for _, legacy := range []bool{true, false} {
+		t.Run(fmt.Sprintf("legacy=%v", legacy), func(t *testing.T) {
+			setMultipartListingTestMode(t, legacy)
+			for range 4 {
+				mp, err := z.NewMultipartUpload(t.Context(), bucket, "valid", ObjectOptions{})
+				if err != nil {
+					t.Fatal(err)
+				}
+				before := notifications.Load()
+				var invalid InvalidUploadID
+				if err := z.AbortMultipartUpload(t.Context(), bucket, "wrong", mp.UploadID, ObjectOptions{}); !errors.As(err, &invalid) {
+					t.Fatalf("wrong target: %v", err)
+				}
+				if notifications.Load() != before {
+					t.Fatal("wrong-target abort sent a destructive peer notification")
+				}
+				if err := z.AbortMultipartUpload(t.Context(), bucket, "valid", mp.UploadID, ObjectOptions{}); err != nil {
+					t.Fatal(err)
+				}
+				if notifications.Load() != before+1 {
+					t.Fatal("successful abort lost peer notification after unlocking")
+				}
+			}
+		})
+	}
+}
+
+func TestMultipartAbortStrictMajority(t *testing.T) {
+	z, set, bucket := multipartListingFixture(t)
+	mp, err := z.NewMultipartUpload(t.Context(), bucket, "a", ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	original := set.getDisks
+	disks := original()
+	set.getDisks = func() []StorageAPI {
+		wrapped := append([]StorageAPI(nil), disks...)
+		wrapped[0] = multipartListingFaultDisk{StorageAPI: disks[0], delete: func(context.Context, string, string, DeleteOptions) error { return errFaultyDisk }}
+		return wrapped
+	}
+	t.Cleanup(func() { set.getDisks = original })
+	if err := z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+		t.Fatalf("ordinary majority cancellation was tightened: %v", err)
+	}
+	// One failed deletion is tolerated in the ordinary majority path. Retrying
+	// after recovery must also clean the now-minority remnant.
+	if _, err := disks[0].ReadVersion(t.Context(), bucket, minioMetaMultipartBucket, set.getUploadIDDir(bucket, "a", mp.UploadID), "", ReadOptions{}); err != nil {
+		t.Fatalf("fault fixture did not leave a remnant: %v", err)
+	}
+	set.getDisks = original
+	if err := z.AbortMultipartUpload(t.Context(), bucket, "a", mp.UploadID, ObjectOptions{}); err != nil {
+		t.Fatal(err)
+	}
+}
@@ -31,6 +31,7 @@ import (
 	"sync"
 	"time"

+	"github.com/google/uuid"
 	"github.com/klauspost/readahead"
 	"github.com/minio/minio-go/v7/pkg/set"
 	"github.com/minio/minio/internal/config/storageclass"
@@ -44,6 +45,15 @@ import (
 	"github.com/pgsty/silo-pkg/v3/sync/errgroup"
 )

+const (
+	multipartMetaBucket = ReservedMetadataPrefixLower + "multipart-v1-bucket"
+	multipartMetaObject = ReservedMetadataPrefixLower + "multipart-v1-object"
+
+	// ponytail: keep scan concurrency fixed until the multipart-list benchmark
+	// establishes a better adaptive limit.
+	multipartMetadataScanConcurrency = 4
+)
+
 func (er erasureObjects) getUploadIDDir(bucket, object, uploadID string) string {
 	uploadUUID := uploadID
 	uploadBytes, err := base64.RawURLEncoding.DecodeString(uploadID)
@@ -251,14 +261,152 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 	})
 }

-// ListMultipartUploads - lists all the pending multipart
-// uploads for a particular object in a bucket.
-//
-// Implements minimal S3 compatible ListMultipartUploads API. We do
-// not support prefix based listing, this is a deliberate attempt
-// towards simplification of multipart APIs.
-// The resulting ListMultipartsInfo structure is unmarshalled directly as XML.
-func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
+func multipartUploadInfo(bucket, object, uploadUUID string, fallback time.Time) MultipartInfo {
+	initiated := fallback
+	if parsed, ok := multipartUploadTime(uploadUUID); ok {
+		initiated = parsed
+	}
+	return MultipartInfo{
+		Bucket:    bucket,
+		Object:    object,
+		UploadID:  base64.RawURLEncoding.EncodeToString(fmt.Appendf(nil, "%s.%s", globalDeploymentID(), uploadUUID)),
+		Initiated: initiated,
+	}
+}
+
+// multipartUploadTime is shared by stored records and continuation markers.
+// The time is part of the immutable upload ID, so a removed marker still
+// identifies the same ordering boundary.
+func multipartUploadTime(uploadUUID string) (time.Time, bool) {
+	if len(uploadUUID) < 38 || uploadUUID[36] != 'x' {
+		return time.Time{}, false
+	}
+	if _, err := uuid.Parse(uploadUUID[:36]); err != nil {
+		return time.Time{}, false
+	}
+	ns, err := strconv.ParseInt(uploadUUID[37:], 10, 64)
+	if err != nil || ns <= 0 || strconv.FormatInt(ns, 10) != uploadUUID[37:] {
+		return time.Time{}, false
+	}
+	return time.Unix(0, ns), true
+}
+
+func multipartMarkerTime(uploadID string) (time.Time, bool) {
+	b, err := base64.RawURLEncoding.DecodeString(uploadID)
+	if err != nil {
+		return time.Time{}, false
+	}
+	_, uploadUUID, ok := strings.Cut(string(b), ".")
+	if !ok {
+		return time.Time{}, false
+	}
+	return multipartUploadTime(uploadUUID)
+}
+
+type multipartListEntry struct {
+	upload       *MultipartInfo
+	commonPrefix string
+}
+
+// paginateMultipartUploads applies the S3 ordering, prefix, delimiter, marker,
+// and page rules exactly once after all pools and sets have been merged.
+func paginateMultipartUploads(uploads []MultipartInfo, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) ListMultipartsInfo {
+	if maxUploads > maxUploadsList {
+		maxUploads = maxUploadsList
+	}
+	result := ListMultipartsInfo{
+		MaxUploads:     maxUploads,
+		KeyMarker:      keyMarker,
+		UploadIDMarker: uploadIDMarker,
+		Prefix:         prefix,
+		Delimiter:      delimiter,
+	}
+
+	deduplicated := make([]MultipartInfo, 0, len(uploads))
+	seenUploads := make(map[string]struct{}, len(uploads))
+	for _, upload := range uploads {
+		identity := upload.Bucket + "\x00" + upload.Object + "\x00" + upload.UploadID
+		if _, ok := seenUploads[identity]; ok {
+			continue
+		}
+		seenUploads[identity] = struct{}{}
+		deduplicated = append(deduplicated, upload)
+	}
+	sort.Slice(deduplicated, func(i, j int) bool {
+		if deduplicated[i].Object != deduplicated[j].Object {
+			return deduplicated[i].Object < deduplicated[j].Object
+		}
+		if !deduplicated[i].Initiated.Equal(deduplicated[j].Initiated) {
+			return deduplicated[i].Initiated.Before(deduplicated[j].Initiated)
+		}
+		return deduplicated[i].UploadID < deduplicated[j].UploadID
+	})
+
+	markerTime, _ := multipartMarkerTime(uploadIDMarker)
+	seenPrefixes := make(map[string]struct{})
+	entries := make([]multipartListEntry, 0, len(deduplicated))
+	for i := range deduplicated {
+		upload := &deduplicated[i]
+		if !strings.HasPrefix(upload.Object, prefix) {
+			continue
+		}
+		if keyMarker != "" {
+			switch strings.Compare(upload.Object, keyMarker) {
+			case -1:
+				continue
+			case 0:
+				if uploadIDMarker == "" || upload.Initiated.Before(markerTime) ||
+					(upload.Initiated.Equal(markerTime) && upload.UploadID <= uploadIDMarker) {
+					continue
+				}
+			}
+		}
+
+		if delimiter != "" {
+			remainder := strings.TrimPrefix(upload.Object, prefix)
+			if i := strings.Index(remainder, delimiter); i >= 0 {
+				commonPrefix := prefix + remainder[:i+len(delimiter)]
+				if keyMarker != "" && commonPrefix <= keyMarker {
+					continue
+				}
+				if _, ok := seenPrefixes[commonPrefix]; ok {
+					continue
+				}
+				seenPrefixes[commonPrefix] = struct{}{}
+				entries = append(entries, multipartListEntry{commonPrefix: commonPrefix})
+				continue
+			}
+		}
+		entries = append(entries, multipartListEntry{upload: upload})
+	}
+
+	if maxUploads <= 0 {
+		return result
+	}
+	pageSize := min(maxUploads, len(entries))
+	for _, entry := range entries[:pageSize] {
+		if entry.upload != nil {
+			result.Uploads = append(result.Uploads, *entry.upload)
+			continue
+		}
+		result.CommonPrefixes = append(result.CommonPrefixes, entry.commonPrefix)
+	}
+	result.IsTruncated = pageSize < len(entries)
+	if result.IsTruncated && pageSize > 0 {
+		last := entries[pageSize-1]
+		if last.upload != nil {
+			result.NextKeyMarker = last.upload.Object
+			result.NextUploadIDMarker = last.upload.UploadID
+		} else {
+			result.NextKeyMarker = last.commonPrefix
+		}
+	}
+	return result
+}
+
+// listMultipartUploadsExact preserves the hashed exact-object lookup used by
+// multipart write placement and by rolling-upgrade legacy mode.
+func (er erasureObjects) listMultipartUploadsExact(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
 	auditObjectErasureSet(ctx, "ListMultipartUploads", object, &er)

 	result.MaxUploads = maxUploads
@@ -311,20 +459,16 @@ func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, objec
 		if populatedUploadIDs.Contains(uploadID) {
 			continue
 		}
-		// If present, use time stored in ID.
-		startTime := time.Now()
-		if split := strings.Split(uploadID, "x"); len(split) == 2 {
-			t, err := strconv.ParseInt(split[1], 10, 64)
-			if err == nil {
-				startTime = time.Unix(0, t)
+		var fallback time.Time
+		if _, ok := multipartUploadTime(uploadID); !ok {
+			fi, err := disk.ReadVersion(ctx, bucket, minioMetaMultipartBucket,
+				pathJoin(er.getMultipartSHADir(bucket, object), uploadID), "", ReadOptions{})
+			if err != nil {
+				return result, toObjectErr(err, bucket, object)
 			}
+			fallback = fi.ModTime
 		}
-		uploads = append(uploads, MultipartInfo{
-			Bucket:    bucket,
-			Object:    object,
-			UploadID:  base64.RawURLEncoding.EncodeToString(fmt.Appendf(nil, "%s.%s", globalDeploymentID(), uploadID)),
-			Initiated: startTime,
-		})
+		uploads = append(uploads, multipartUploadInfo(bucket, object, uploadID, fallback))
 		populatedUploadIDs.Add(uploadID)
 	}

@@ -365,6 +509,25 @@ func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, objec
 	return result, nil
 }

+func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (ListMultipartsInfo, error) {
+	if err := checkListMultipartArgs(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter); err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	scan, err := startMultipartScan(ctx, false)
+	if err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	defer scan.close()
+	uploads, legacy, err := er.scanMultipartUploads(scan, bucket, 0, 0)
+	if err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	if legacy {
+		return ListMultipartsInfo{}, errMultipartListingLegacy
+	}
+	return paginateMultipartUploads(uploads, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads), nil
+}
+
 // newMultipartUpload - wrapper for initializing a new multipart
 // request; returns a unique upload id.
 //
@@ -402,6 +565,8 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 	}

 	userDefined := cloneMSS(opts.UserDefined)
+	userDefined[multipartMetaBucket] = bucket
+	userDefined[multipartMetaObject] = object
 	if opts.PreserveETag != "" {
 		userDefined["etag"] = opts.PreserveETag
 	}
@@ -1459,6 +1624,8 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	// Remove superfluous internal headers.
 	delete(fi.Metadata, hash.MinIOMultipartChecksum)
 	delete(fi.Metadata, hash.MinIOMultipartChecksumType)
+	delete(fi.Metadata, multipartMetaBucket)
+	delete(fi.Metadata, multipartMetaObject)

 	// Save the final object size and modtime.
 	fi.Size = objectSize
@@ -1586,22 +1753,91 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	return fi.ToObjectInfo(bucket, object, opts.Versioned || opts.VersionSuspended), nil
 }

-// AbortMultipartUpload - aborts an ongoing multipart operation
-// signified by the input uploadID. This is an atomic operation
-// doesn't require clients to initiate multiple such requests.
-//
-// All parts are purged from all disks and reference to the uploadID
-// would be removed from the system, rollback is not possible on this
-// operation.
-func (er erasureObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (err error) {
+// abortMultipartUpload retains read-quorum validation and best-effort cleanup
+// in legacy mode. Strict mode requires majority deletion acknowledgements and
+// permits retrying remnants below read quorum. Neither mode fences creation
+// writes still executing after a storage timeout.
+func (er erasureObjects) abortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions, legacy bool) (bool, error) {
 	if !opts.NoAuditLog {
 		auditObjectErasureSet(ctx, "AbortMultipartUpload", object, &er)
 	}
-
-	// Cleanup all uploaded parts.
-	defer er.deleteAll(ctx, minioMetaMultipartBucket, er.getUploadIDDir(bucket, object, uploadID))
-
-	// Validates if upload ID exists.
-	_, _, err = er.checkUploadIDExists(ctx, bucket, object, uploadID, false)
-	return toObjectErr(err, bucket, object, uploadID)
+	b, err := base64.RawURLEncoding.DecodeString(uploadID)
+	if err != nil {
+		return false, MalformedUploadID{UploadID: uploadID}
+	}
+	_, internalID, ok := strings.Cut(string(b), ".")
+	if !ok || internalID == "" || internalID == "." || internalID == ".." || strings.ContainsAny(internalID, "/\\") {
+		return false, InvalidUploadID{Bucket: bucket, Object: object, UploadID: uploadID}
+	}
+	if legacy {
+		// Keep the released read-quorum and best-effort cleanup behavior.
+		// The upload ID safety check above applies to both modes.
+		defer er.deleteAll(ctx, minioMetaMultipartBucket, er.getUploadIDDir(bucket, object, uploadID))
+		_, _, err := er.checkUploadIDExists(ctx, bucket, object, uploadID, false)
+		err = toObjectErr(err, bucket, object, uploadID)
+		if _, absent := err.(InvalidUploadID); absent {
+			return false, nil
+		}
+		return err == nil, err
+	}
+	disks := er.getDisks()
+	uploadPath := er.getUploadIDDir(bucket, object, uploadID)
+	_, errs := readAllFileInfo(ctx, disks, bucket, minioMetaMultipartBucket, uploadPath, "", false, false)
+	quorum := er.setDriveCount/2 + 1
+	found, absent := false, 0
+	observed := make([]bool, len(disks))
+	for i, err := range errs {
+		switch {
+		case err == nil, errors.Is(err, errFileCorrupt):
+			found = true
+			observed[i] = true
+		case errors.Is(err, errFileNotFound), errors.Is(err, errFileVersionNotFound):
+			absent++
+		}
+	}
+	if absent >= quorum && !found {
+		return false, nil
+	}
+	if !found {
+		return false, toObjectErr(errErasureReadQuorum, bucket, object, uploadID)
+	}
+	g := errgroup.WithNErrs(len(disks))
+	for i, disk := range disks {
+		g.Go(func() error {
+			if disk == nil {
+				return errDiskNotFound
+			}
+			err := disk.Delete(ctx, minioMetaMultipartBucket, uploadPath, DeleteOptions{Recursive: true, Immediate: false})
+			if errors.Is(err, errFileNotFound) || errors.Is(err, errFileVersionNotFound) {
+				return nil
+			}
+			return err
+		}, i)
+	}
+	deleteErrs := g.Wait()
+	if err := reduceWriteQuorumErrs(ctx, deleteErrs, nil, quorum); err != nil {
+		return true, toObjectErr(err, bucket, object, uploadID)
+	}
+	if absent >= quorum {
+		// Missing disks must not mask a failed cleanup of known remnants.
+		for i, found := range observed {
+			if found && deleteErrs[i] != nil {
+				return true, toObjectErr(errErasureWriteQuorum, bucket, object, uploadID)
+			}
+		}
+	}
+	return true, nil
+}
+
+// AbortMultipartUpload cancels an upload using the configured mode. Offline
+// part data may still need stale-upload cleanup after its drives return.
+func (er erasureObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (err error) {
+	found, err := er.abortMultipartUpload(ctx, bucket, object, uploadID, opts, globalAPIConfig.getMultipartListingLegacy())
+	if err != nil {
+		return err
+	}
+	if !found {
+		return InvalidUploadID{Bucket: bucket, Object: object, UploadID: uploadID}
+	}
+	return nil
 }
@@ -1657,7 +1657,7 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	return fi.ToObjectInfo(bucket, object, opts.Versioned || opts.VersionSuspended), nil
 }

-func (er erasureObjects) deleteObjectVersion(ctx context.Context, bucket, object string, fi FileInfo, forceDelMarker bool) error {
+func (er erasureObjects) deleteObjectVersion(ctx context.Context, bucket, object string, fi FileInfo, forceDelMarker, purge bool) error {
 	disks := er.getDisks()
 	// Assume (N/2 + 1) quorum for Delete()
 	// this is a theoretical assumption such that
@@ -1673,7 +1673,13 @@ func (er erasureObjects) deleteObjectVersion(ctx context.Context, bucket, object
 			if disks[index] == nil {
 				return errDiskNotFound
 			}
-			return disks[index].DeleteVersion(ctx, bucket, object, fi, forceDelMarker, DeleteOptions{})
+			err := disks[index].DeleteVersion(ctx, bucket, object, fi, forceDelMarker, DeleteOptions{})
+			// Physical removal and reliable already-absent replies are the same
+			// outcome. Creation and metadata updates must not use this quorum.
+			if purge && (err == errFileNotFound || err == errFileVersionNotFound) {
+				return nil
+			}
+			return err
 		}, index)
 	}
 	// return errors if any during deletion
@@ -2023,6 +2029,11 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 		if opts.DeleteMarker {
 			versionFound = false
 		} else if !tryDel {
+			if opts.isVersionPurge() && (isErrObjectNotFound(gerr) || isErrVersionNotFound(gerr)) {
+				if err := er.checkPurgeAbsent(ctx, bucket, object, opts.VersionID); err != nil {
+					return objInfo, err
+				}
+			}
 			return objInfo, gerr
 		}
 	}
@@ -2105,6 +2116,13 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 		}
 	}

+	purge := opts.isVersionPurge()
+	if purge {
+		// A delete marker describes the version being removed; it is not an
+		// instruction to create that marker on disks which already lack it.
+		markDelete, deleteMarker = false, false
+	}
+
 	modTime := opts.MTime
 	if opts.MTime.IsZero() {
 		modTime = UTCNow()
@@ -2151,7 +2169,7 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 		// delete marker. Add delete marker, since we don't have
 		// any version specified explicitly. Or if a particular
 		// version id needs to be replicated.
-		if err = er.deleteObjectVersion(ctx, bucket, object, fi, opts.DeleteMarker); err != nil {
+		if err = er.deleteObjectVersion(ctx, bucket, object, fi, opts.DeleteMarker, false); err != nil {
 			return objInfo, toObjectErr(err, bucket, object)
 		}
 		oi := fi.ToObjectInfo(bucket, object, opts.Versioned || opts.VersionSuspended)
@@ -2174,11 +2192,19 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 	if opts.SkipFreeVersion {
 		dfi.SetSkipTierFreeVersion()
 	}
-	if err = er.deleteObjectVersion(ctx, bucket, object, dfi, opts.DeleteMarker); err != nil {
+	if err = er.deleteObjectVersion(ctx, bucket, object, dfi, opts.DeleteMarker, purge); err != nil {
 		return objInfo, toObjectErr(err, bucket, object)
 	}

-	return dfi.ToObjectInfo(bucket, object, opts.Versioned || opts.VersionSuspended), nil
+	oi := dfi.ToObjectInfo(bucket, object, opts.Versioned || opts.VersionSuspended)
+	if purge {
+		// Preserve the DELETE response's identity without reusing it as a disk
+		// instruction to create a marker. The lookup also exposes a data
+		// version pending purge as deleted for visibility; only a stored
+		// marker, which carries no erasure layout, is reported as one.
+		oi.DeleteMarker = goi.DeleteMarker && goi.DataBlocks == 0
+	}
+	return oi, nil
 }

 // Send the successful but partial upload/delete, however ignore
@@ -2467,7 +2493,7 @@ func (er erasureObjects) TransitionObject(ctx context.Context, bucket, object st

 	storageDisks := er.getDisks()

-	if err = er.deleteObjectVersion(ctx, bucket, object, fi, false); err != nil {
+	if err = er.deleteObjectVersion(ctx, bucket, object, fi, false, false); err != nil {
 		eventName = event.ObjectTransitionFailed
 	}

@@ -316,6 +316,11 @@ func (z *erasureServerPools) retireReplicaCopies(ctx context.Context, bucket, ob
 func (z *erasureServerPools) deleteObjectReconciled(ctx context.Context, bucket, object string, opts ObjectOptions) (ObjectInfo, error) {
 	copies, err := z.objectPoolInfos(ctx, bucket, object, opts)
 	if err != nil {
+		if opts.isVersionPurge() && (isErrObjectNotFound(err) || isErrVersionNotFound(err)) {
+			if quorumErr := z.checkPurgeAbsent(ctx, bucket, object, opts.VersionID); quorumErr != nil {
+				return ObjectInfo{}, quorumErr
+			}
+		}
 		return ObjectInfo{}, err
 	}
 	primary := copies[0]
@@ -365,6 +370,21 @@ func (z *erasureServerPools) deleteObjectReconciled(ctx context.Context, bucket,
 			opts.EvalMetadataFn = nil
 		}
 	}
+	if opts.isVersionPurge() {
+		// A missing pool may have only read-quorum absence. Verify every pool
+		// omitted by lookup before mutating the known copies.
+		for i, pool := range z.serverPools {
+			found := false
+			for _, copy := range copies {
+				found = found || copy.Index == i
+			}
+			if !found {
+				if err := pool.getHashedSet(object).checkPurgeAbsent(ctx, bucket, object, opts.VersionID); err != nil {
+					return ObjectInfo{}, err
+				}
+			}
+		}
+	}
 	if opts.VersionID == "" && (opts.Versioned || opts.VersionSuspended) {
 		// A single new delete marker hides the current version. Older versions
 		// remain history and must not be removed by an unqualified DELETE.
@@ -618,7 +618,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, idx int, bu
 	if objInfo.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, objInfo.Name, ObjectOptions{
 			VersionID:    objInfo.VersionID,
-			UserDefined:  objInfo.UserDefined,
+			UserDefined:  migrationObjectMetadata(objInfo),
 			NoAuditLog:   true,
 			SrcPoolIdx:   idx,
 			DataMovement: true,
@@ -681,7 +681,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, idx int, bu
 			SrcPoolIdx:   idx,
 			VersionID:    objInfo.VersionID,
 			MTime:        objInfo.ModTime,
-			UserDefined:  objInfo.UserDefined,
+			UserDefined:  migrationObjectMetadata(objInfo),
 			PreserveETag: objInfo.ETag, // Preserve original ETag to ensure same metadata.
 			IndexCB: func() []byte {
 				return objInfo.Parts[0].Index // Preserve part Index to ensure decompression works.
@@ -0,0 +1,33 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import xhttp "github.com/minio/minio/internal/http"
+
+// migrationObjectMetadata restores tags removed from the read representation.
+// Keep the original revision, including ordered empty states, for the existing
+// locked reconciliation at PUT or multipart completion. Moving is not a new
+// tagging mutation, and the write must not modify the reader's metadata map.
+func migrationObjectMetadata(oi ObjectInfo) map[string]string {
+	metadata := cloneMSS(oi.UserDefined)
+	delete(metadata, xhttp.AmzObjectTagging)
+	if oi.UserTags != "" {
+		metadata[xhttp.AmzObjectTagging] = oi.UserTags
+	}
+	return metadata
+}
@@ -0,0 +1,592 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"maps"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	xhttp "github.com/minio/minio/internal/http"
+)
+
+const migrationTagRevision = ReservedMetadataPrefixLower + TaggingTimestamp
+
+var migrationTagStates = []struct {
+	name, tags, revision string
+	hasRevision          bool
+}{
+	{name: "initial", tags: "team=storage&path=a%2Fb"},
+	{name: "empty-revision", tags: "team=storage", hasRevision: true},
+	{name: "invalid-revision", tags: "team=storage", revision: "invalid", hasRevision: true},
+	{name: "ordered", tags: "team=storage", revision: "2026-09-16T09:00:00Z", hasRevision: true},
+	{name: "cleared", revision: "2026-09-16T09:00:00Z", hasRevision: true},
+	{name: "never-tagged"},
+	{name: "empty-empty-revision", hasRevision: true},
+	{name: "empty-invalid-revision", revision: "invalid", hasRevision: true},
+}
+
+func TestMigrationObjectMetadata(t *testing.T) {
+	for _, state := range migrationTagStates {
+		for _, residual := range []bool{false, true} {
+			t.Run(fmt.Sprintf("%s/residual=%t", state.name, residual), func(t *testing.T) {
+				raw := map[string]string{
+					xhttp.AmzObjectTagging:                    state.tags,
+					"x-amz-meta-team":                         "storage",
+					"x-amz-meta-empty":                        "",
+					ReservedMetadataPrefix + "compression":    "opaque-compression-state",
+					ReservedMetadataPrefix + "sealed-key":     "opaque-key-state",
+					"x-amz-object-lock-retain-until-date":     "2030-01-01T00:00:00Z",
+					"x-amz-object-lock-legal-hold":            "ON",
+					"x-amz-tagging":                           "keep-noncanonical-metadata",
+					ReservedMetadataPrefix + "actual-size":    "42",
+					ReservedMetadataPrefix + "replica-status": "opaque-replication-state",
+				}
+				if state.hasRevision {
+					raw[migrationTagRevision] = state.revision
+				}
+				oi := (FileInfo{Metadata: raw}).ToObjectInfo("bucket", "object", false)
+				if residual {
+					oi.UserDefined[xhttp.AmzObjectTagging] = "stale=raw"
+				}
+				before := maps.Clone(oi.UserDefined)
+				got := migrationObjectMetadata(oi)
+				if got == nil || !maps.Equal(before, oi.UserDefined) {
+					t.Fatal("helper must return a new non-nil map without changing the read snapshot")
+				}
+				value, exists := got[xhttp.AmzObjectTagging]
+				if value != state.tags || exists != (state.tags != "") {
+					t.Fatalf("raw tags=(%q,%t), want (%q,%t)", value, exists, state.tags, state.tags != "")
+				}
+				stamp, present := got[migrationTagRevision]
+				if stamp != state.revision || present != state.hasRevision {
+					t.Fatalf("revision=(%q,%t), want (%q,%t)", stamp, present, state.revision, state.hasRevision)
+				}
+				for key, value := range before {
+					if stored, exists := got[key]; key != xhttp.AmzObjectTagging && (!exists || stored != value) {
+						t.Errorf("lost metadata %q", key)
+					}
+				}
+				got["x-amz-meta-team"] = "changed"
+				if !maps.Equal(before, oi.UserDefined) {
+					t.Fatal("output aliases the read snapshot")
+				}
+			})
+		}
+	}
+	for _, metadata := range []map[string]string{nil, {}} {
+		got := migrationObjectMetadata(ObjectInfo{UserDefined: metadata})
+		if got == nil || len(got) != 0 {
+			t.Fatalf("empty input must yield a writable empty map: %#v", got)
+		}
+		got["new"] = "value"
+		if len(metadata) != 0 {
+			t.Fatal("empty map input was aliased")
+		}
+	}
+}
+
+// Allocation ignores the host's used-space percentage, as in the existing tag
+// fixtures. All object data and metadata still use real erasure-storage disks.
+func migrationTestPools(t *testing.T) (*erasureServerPools, string) {
+	t.Helper()
+	z, bucket := consistencyPools(t)
+	for _, pool := range z.serverPools {
+		for _, set := range pool.sets {
+			original := set.getDisks
+			disks := append([]StorageAPI(nil), original()...)
+			for i := range disks {
+				disks[i] = tagTestCapacityDisk{StorageAPI: disks[i]}
+			}
+			set.getDisks = func() []StorageAPI { return disks }
+			t.Cleanup(func() { set.getDisks = original })
+		}
+	}
+	return z, bucket
+}
+
+func migrationTestMover(t *testing.T, z *erasureServerPools, source int, kind string) func(context.Context, int, string, *GetObjectReader) error {
+	t.Helper()
+	if kind == "rebalance" {
+		z.rebalMu.Lock()
+		z.rebalMeta = &rebalanceMeta{PoolStats: []*rebalanceStats{{}, {}}}
+		z.rebalMeta.PoolStats[source] = &rebalanceStats{Participating: true, Info: rebalanceInfo{Status: rebalStarted}}
+		z.rebalMu.Unlock()
+		t.Cleanup(func() { z.rebalMu.Lock(); z.rebalMeta = nil; z.rebalMu.Unlock() })
+		return z.rebalanceObject
+	}
+	z.poolMetaMutex.Lock()
+	z.poolMeta.Pools[source].Decommission = &PoolDecommissionInfo{}
+	z.poolMetaMutex.Unlock()
+	t.Cleanup(func() { z.poolMetaMutex.Lock(); z.poolMeta.Pools[source].Decommission = nil; z.poolMetaMutex.Unlock() })
+	return z.decommissionObject
+}
+
+func migrationTestObject(t *testing.T, z *erasureServerPools, bucket, object string, source int, multipart bool, parts []string, opts ObjectOptions) ObjectInfo {
+	t.Helper()
+	if !multipart {
+		oi := putConsistencyObject(t, z, bucket, object, source, strings.Join(parts, ""), opts)
+		return migrationTestPersistedInfo(t, z, source, oi)
+	}
+	mp, err := z.serverPools[source].NewMultipartUpload(t.Context(), bucket, object, opts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	completed := make([]CompletePart, len(parts))
+	for i, body := range parts {
+		part, err := z.serverPools[source].PutObjectPart(t.Context(), bucket, object, mp.UploadID, i+1,
+			mustGetPutObjReader(t, strings.NewReader(body), int64(len(body)), "", ""), ObjectOptions{})
+		if err != nil {
+			t.Fatal(err)
+		}
+		completed[i] = CompletePart{PartNumber: i + 1, ETag: part.ETag}
+	}
+	oi, err := z.serverPools[source].CompleteMultipartUpload(t.Context(), bucket, object, mp.UploadID, completed, opts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !oi.isMultipart() {
+		t.Fatal("fixture did not create a multipart object")
+	}
+	return migrationTestPersistedInfo(t, z, source, oi)
+}
+
+func migrationTestPersistedInfo(t *testing.T, z *erasureServerPools, pool int, oi ObjectInfo) ObjectInfo {
+	t.Helper()
+	// PUT can return transient fields such as tier-free-versionID that are not
+	// stored. Take the same persisted read representation used by the movers.
+	got, err := z.serverPools[pool].GetObjectInfo(t.Context(), oi.Bucket, oi.Name, ObjectOptions{VersionID: migrationTestVersion(oi)})
+	if err != nil {
+		t.Fatal(err)
+	}
+	return got
+}
+
+func migrationTestVersion(oi ObjectInfo) string {
+	if oi.VersionID == "" {
+		return nullVersionID
+	}
+	return oi.VersionID
+}
+
+func migrationTestReader(t *testing.T, z *erasureServerPools, source int, oi ObjectInfo) *GetObjectReader {
+	t.Helper()
+	gr, err := z.serverPools[source].GetObjectNInfo(t.Context(), oi.Bucket, oi.Name, nil, nil,
+		ObjectOptions{VersionID: migrationTestVersion(oi), NoLock: true, NoDecryption: true})
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { gr.Close() })
+	return gr
+}
+
+func migrationTestStored(t *testing.T, z *erasureServerPools, pool int, original ObjectInfo, tags, revision string, hasRevision bool, body string) {
+	t.Helper()
+	version := migrationTestVersion(original)
+	got, err := z.serverPools[pool].GetObjectInfo(t.Context(), original.Bucket, original.Name, ObjectOptions{VersionID: version})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got.UserTags != tags || got.ETag != original.ETag || !got.ModTime.Equal(original.ModTime) || got.VersionID != original.VersionID || got.Size != original.Size {
+		t.Errorf("pool %d changed tags/object identity: tags=%q want=%q version=%q/%q etag=%q/%q size=%d/%d mtime=%v/%v", pool, got.UserTags, tags, got.VersionID, original.VersionID, got.ETag, original.ETag, got.Size, original.Size, got.ModTime, original.ModTime)
+	}
+	infos, errs := readAllFileInfo(t.Context(), z.serverPools[pool].getHashedSet(original.Name).getDisks(), "", original.Bucket, original.Name, version, false, false)
+	for disk, fi := range infos {
+		if errs[disk] != nil {
+			t.Fatalf("pool %d disk %d: %v", pool, disk, errs[disk])
+		}
+		stamp, exists := fi.Metadata[migrationTagRevision]
+		if fi.Metadata[xhttp.AmzObjectTagging] != tags || stamp != revision || exists != hasRevision {
+			t.Errorf("pool %d disk %d: tags=%q revision=(%q,%t), want %q (%q,%t)", pool, disk, fi.Metadata[xhttp.AmzObjectTagging], stamp, exists, tags, revision, hasRevision)
+		}
+		for key, value := range original.UserDefined {
+			if key != migrationTagRevision && fi.Metadata[key] != value {
+				t.Errorf("pool %d disk %d lost metadata %q", pool, disk, key)
+			}
+		}
+	}
+	gr, err := z.serverPools[pool].GetObjectNInfo(t.Context(), original.Bucket, original.Name, nil, nil, ObjectOptions{VersionID: version})
+	if err != nil {
+		t.Fatal(err)
+	}
+	data, err := io.ReadAll(gr)
+	gr.Close()
+	if err != nil || !bytes.Equal(data, []byte(body)) {
+		t.Fatalf("pool %d content changed: bytes=%d err=%v", pool, len(data), err)
+	}
+}
+
+func TestPoolsMigrationPreservesTagState(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for _, method := range []string{"put", "multipart"} {
+			for source := range 2 {
+				for _, state := range migrationTagStates {
+					t.Run(fmt.Sprintf("%s/%s/source=%d/%s", kind, method, source, state.name), func(t *testing.T) {
+						move := migrationTestMover(t, z, source, kind)
+						meta := map[string]string{xhttp.AmzObjectTagging: state.tags, "x-amz-meta-owner": "retained"}
+						if state.hasRevision {
+							meta[migrationTagRevision] = state.revision
+						}
+						oi := migrationTestObject(t, z, bucket, t.Name(), source, method == "multipart", []string{"payload"}, ObjectOptions{Versioned: true, UserDefined: meta})
+						migrationTestStored(t, z, source, oi, state.tags, state.revision, state.hasRevision, "payload")
+						gr := migrationTestReader(t, z, source, oi)
+						before := maps.Clone(gr.ObjInfo.UserDefined)
+						if err := move(t.Context(), source, bucket, gr); err != nil {
+							t.Fatal(err)
+						}
+						if !maps.Equal(before, gr.ObjInfo.UserDefined) {
+							t.Error("migration modified its source snapshot")
+						}
+						migrationTestStored(t, z, 1-source, oi, state.tags, state.revision, state.hasRevision, "payload")
+					})
+				}
+			}
+		}
+	}
+}
+
+type migrationTestGate struct {
+	entered, resume chan struct{}
+	once, release   sync.Once
+}
+
+func newMigrationTestGate() *migrationTestGate {
+	return &migrationTestGate{entered: make(chan struct{}), resume: make(chan struct{})}
+}
+
+func (g *migrationTestGate) wait(ctx context.Context) error {
+	g.once.Do(func() { close(g.entered) })
+	select {
+	case <-g.resume:
+		return nil
+	case <-ctx.Done():
+		return ctx.Err()
+	}
+}
+
+func (g *migrationTestGate) unblock() { g.release.Do(func() { close(g.resume) }) }
+
+type migrationTestGateReader struct {
+	io.Reader
+	ctx  context.Context
+	gate *migrationTestGate
+}
+
+func (r migrationTestGateReader) Read(p []byte) (int, error) {
+	if err := r.gate.wait(r.ctx); err != nil {
+		return 0, err
+	}
+	return r.Reader.Read(p)
+}
+
+func TestPoolsMigrationRechecksTags(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	const recent = "2026-09-16T10:00:00Z"
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for _, method := range []string{"put", "multipart"} {
+			for source := range 2 {
+				for _, tags := range []string{"state=after", ""} {
+					t.Run(fmt.Sprintf("%s/%s/source=%d/clear=%t", kind, method, source, tags == ""), func(t *testing.T) {
+						move := migrationTestMover(t, z, source, kind)
+						oi := migrationTestObject(t, z, bucket, t.Name(), source, method == "multipart", []string{"payload"}, ObjectOptions{
+							Versioned: true, UserDefined: map[string]string{xhttp.AmzObjectTagging: "state=before"},
+						})
+						gr := migrationTestReader(t, z, source, oi)
+						before := maps.Clone(gr.ObjInfo.UserDefined)
+						ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+						defer cancel()
+						gate := newMigrationTestGate()
+						done, finished := make(chan error, 1), make(chan struct{})
+						if method == "multipart" {
+							// The first part read is after persisted upload initialization,
+							// but before completion takes the object lock and reconciles.
+							gr.Reader = migrationTestGateReader{Reader: gr.Reader, ctx: ctx, gate: gate}
+							go func() { defer close(finished); done <- move(ctx, source, bucket, gr) }()
+							defer func() { gate.unblock(); cancel(); <-finished }()
+							select {
+							case <-gate.entered:
+							case err := <-done:
+								t.Fatalf("migration finished before the part barrier: %v", err)
+							case <-ctx.Done():
+								t.Fatal(ctx.Err())
+							}
+							uploads, err := z.serverPools[1-source].listMultipartUploadsExact(ctx, bucket, oi.Name)
+							if err != nil || len(uploads.Uploads) != 1 {
+								t.Fatalf("upload not persisted at barrier: %+v, %v", uploads, err)
+							}
+							info, err := z.serverPools[1-source].GetMultipartInfo(ctx, bucket, oi.Name, uploads.Uploads[0].UploadID, ObjectOptions{})
+							if err != nil || info.UserDefined[xhttp.AmzObjectTagging] != "state=before" {
+								t.Fatalf("upload lost snapshot tags: %v, %v", info.UserDefined, err)
+							}
+						}
+						// For ordinary PUT this must finish before calling the mover:
+						// blocking its data Read would already hold the object lock.
+						opts := ObjectOptions{VersionID: migrationTestVersion(oi), UserDefined: map[string]string{migrationTagRevision: recent}}
+						var err error
+						if tags == "" {
+							_, err = z.DeleteObjectTags(ctx, bucket, oi.Name, opts)
+						} else {
+							_, err = z.PutObjectTags(ctx, bucket, oi.Name, tags, opts)
+						}
+						if err != nil {
+							t.Fatal(err)
+						}
+						if gr.ObjInfo.UserTags != "state=before" || !maps.Equal(before, gr.ObjInfo.UserDefined) {
+							t.Fatal("test must retain the old source snapshot")
+						}
+						migrationTestStored(t, z, source, oi, tags, recent, true, "payload")
+						if method == "multipart" {
+							gate.unblock()
+							err = <-done
+						} else {
+							err = move(ctx, source, bucket, gr)
+						}
+						if err != nil {
+							t.Fatal(err)
+						}
+						if !maps.Equal(before, gr.ObjInfo.UserDefined) {
+							t.Error("reconciliation changed the reader's map")
+						}
+						migrationTestStored(t, z, 1-source, oi, tags, recent, true, "payload")
+					})
+				}
+			}
+		}
+	}
+}
+
+type migrationTestMetadataGateDisk struct {
+	StorageAPI
+	bucket, object string
+	gate           *migrationTestGate
+}
+
+func (d migrationTestMetadataGateDisk) UpdateMetadata(ctx context.Context, volume, path string, fi FileInfo, opts UpdateMetadataOpts) error {
+	if volume == d.bucket && path == d.object {
+		if err := d.gate.wait(ctx); err != nil {
+			return err
+		}
+	}
+	return d.StorageAPI.UpdateMetadata(ctx, volume, path, fi, opts)
+}
+
+func TestPoolsMigrationTagsDuringCleanup(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	const recent = "2026-09-16T10:00:00Z"
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for source := range 2 {
+			for _, tags := range []string{"state=after", ""} {
+				for _, interrupt := range []bool{false, true} {
+					t.Run(fmt.Sprintf("%s/source=%d/clear=%t/interrupt=%t", kind, source, tags == "", interrupt), func(t *testing.T) {
+						move := migrationTestMover(t, z, source, kind)
+						oi := migrationTestObject(t, z, bucket, t.Name(), source, false, []string{"payload"}, ObjectOptions{
+							Versioned: true, UserDefined: map[string]string{xhttp.AmzObjectTagging: "state=before"},
+						})
+						if err := move(t.Context(), source, bucket, migrationTestReader(t, z, source, oi)); err != nil {
+							t.Fatal(err)
+						}
+						migrationTestStored(t, z, 1-source, oi, "state=before", "", false, "payload")
+						ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+						defer cancel()
+						mutate := func() (ObjectInfo, error) {
+							opts := ObjectOptions{VersionID: migrationTestVersion(oi), UserDefined: map[string]string{migrationTagRevision: recent}}
+							if tags == "" {
+								return z.DeleteObjectTags(ctx, bucket, oi.Name, opts)
+							}
+							return z.PutObjectTags(ctx, bucket, oi.Name, tags, opts)
+						}
+						set := z.serverPools[source].getHashedSet(oi.Name)
+						cleanup := func() {
+							// Use the same source prefix cleanup as both outer movers.
+							if _, err := set.DeleteObject(ctx, bucket, oi.Name, ObjectOptions{DeletePrefix: true, DeletePrefixObject: true}); err != nil {
+								t.Fatal(err)
+							}
+						}
+						var updated ObjectInfo
+						if !interrupt {
+							var err error
+							updated, err = mutate()
+							if err != nil {
+								t.Fatal(err)
+							}
+							cleanup()
+						} else {
+							gate := newMigrationTestGate()
+							original := set.getDisks
+							disks := append([]StorageAPI(nil), original()...)
+							for i := range disks {
+								disks[i] = migrationTestMetadataGateDisk{StorageAPI: disks[i], bucket: bucket, object: oi.Name, gate: gate}
+							}
+							set.getDisks = func() []StorageAPI { return disks }
+							defer func() { set.getDisks = original }()
+							done, finished := make(chan error, 1), make(chan struct{})
+							go func() { defer close(finished); _, err := mutate(); done <- err }()
+							defer func() { gate.unblock(); cancel(); <-finished }()
+							select {
+							case <-gate.entered:
+							case err := <-done:
+								t.Fatalf("metadata update missed the barrier: %v", err)
+							case <-ctx.Done():
+								t.Fatal(ctx.Err())
+							}
+							cleanup()
+							gate.unblock()
+							if err := <-done; err == nil {
+								t.Fatal("update reported success after its source copy was removed")
+							}
+							<-finished
+							set.getDisks = original
+							var err error
+							updated, err = mutate()
+							if err != nil {
+								t.Fatalf("metadata retry did not converge: %v", err)
+							}
+						}
+						if _, err := z.serverPools[source].GetObjectInfo(ctx, bucket, oi.Name, ObjectOptions{VersionID: oi.VersionID}); !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+							t.Fatalf("source cleanup left a copy: %v", err)
+						}
+						migrationTestStored(t, z, 1-source, oi, tags, updated.UserDefined[migrationTagRevision], true, "payload")
+					})
+				}
+			}
+		}
+	}
+}
+
+func TestPoolsMigrationVersionHistory(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for _, method := range []string{"put", "multipart"} {
+			for _, versioning := range []string{"unversioned", "null", "history"} {
+				t.Run(kind+"/"+method+"/"+versioning, func(t *testing.T) {
+					move := migrationTestMover(t, z, 0, kind)
+					opts := ObjectOptions{
+						Versioned: versioning == "history", VersionSuspended: versioning == "null",
+						MTime:       time.Date(2026, 9, 16, 8, 0, 0, 0, time.UTC),
+						UserDefined: map[string]string{xhttp.AmzObjectTagging: "version=old"},
+					}
+					versions := []ObjectInfo{migrationTestObject(t, z, bucket, t.Name(), 0, method == "multipart", []string{"old"}, opts)}
+					if versioning == "history" {
+						opts.MTime = opts.MTime.Add(time.Minute)
+						versions = append(versions, migrationTestObject(t, z, bucket, t.Name(), 0, method == "multipart", []string{"old"}, opts))
+					}
+					var newest ObjectInfo
+					if versioning != "unversioned" {
+						newest = migrationTestObject(t, z, bucket, t.Name(), 0, false, []string{"latest"}, ObjectOptions{
+							Versioned: true, MTime: opts.MTime.Add(time.Minute),
+							UserDefined: map[string]string{xhttp.AmzObjectTagging: "version=new"},
+						})
+					}
+					for _, oi := range versions {
+						if err := move(t.Context(), 0, bucket, migrationTestReader(t, z, 0, oi)); err != nil {
+							t.Fatal(err)
+						}
+						migrationTestStored(t, z, 1, oi, "version=old", "", false, "old")
+					}
+					if versioning != "unversioned" {
+						migrationTestStored(t, z, 0, newest, "version=new", "", false, "latest")
+						if _, err := z.serverPools[1].GetObjectInfo(t.Context(), bucket, t.Name(), ObjectOptions{VersionID: newest.VersionID}); !isErrVersionNotFound(err) {
+							t.Fatalf("moving an addressed version affected the newer version: %v", err)
+						}
+					}
+				})
+			}
+		}
+	}
+}
+
+func TestPoolsMigrationMultipartParts(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	parts := []string{strings.Repeat("a", 5<<20), "tail-with-tags"}
+	body := strings.Join(parts, "")
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for source := range 2 {
+			t.Run(fmt.Sprintf("%s/source=%d", kind, source), func(t *testing.T) {
+				move := migrationTestMover(t, z, source, kind)
+				oi := migrationTestObject(t, z, bucket, t.Name(), source, true, parts, ObjectOptions{
+					Versioned: true, UserDefined: map[string]string{xhttp.AmzObjectTagging: "multipart=kept"},
+				})
+				if err := move(t.Context(), source, bucket, migrationTestReader(t, z, source, oi)); err != nil {
+					t.Fatal(err)
+				}
+				migrationTestStored(t, z, 1-source, oi, "multipart=kept", "", false, body)
+				got, err := z.serverPools[1-source].GetObjectInfo(t.Context(), bucket, oi.Name, ObjectOptions{VersionID: oi.VersionID})
+				if err != nil || len(got.Parts) != len(oi.Parts) {
+					t.Fatalf("parts changed: %+v, %v", got.Parts, err)
+				}
+				for i, part := range got.Parts {
+					old := oi.Parts[i]
+					if part.Number != old.Number || part.Size != old.Size || part.ActualSize != old.ActualSize || part.ETag != old.ETag || !bytes.Equal(part.Index, old.Index) {
+						t.Errorf("part %d changed: %+v / %+v", i, part, old)
+					}
+				}
+				start := int64(len(parts[0]) - 3)
+				gr, err := z.serverPools[1-source].GetObjectNInfo(t.Context(), bucket, oi.Name, &HTTPRangeSpec{Start: start, End: start + 7}, nil, ObjectOptions{VersionID: oi.VersionID})
+				if err != nil {
+					t.Fatal(err)
+				}
+				data, err := io.ReadAll(gr)
+				gr.Close()
+				if err != nil || string(data) != body[start:start+8] {
+					t.Fatalf("cross-part range changed: %q, %v", data, err)
+				}
+			})
+		}
+	}
+}
+
+func TestPoolsMigrationUnreadablePool(t *testing.T) {
+	z, bucket := migrationTestPools(t)
+	for _, kind := range []string{"rebalance", "decommission"} {
+		for _, method := range []string{"put", "multipart"} {
+			t.Run(kind+"/"+method, func(t *testing.T) {
+				move := migrationTestMover(t, z, 0, kind)
+				oi := migrationTestObject(t, z, bucket, t.Name(), 0, method == "multipart", []string{"payload"}, ObjectOptions{
+					Versioned: true, UserDefined: map[string]string{xhttp.AmzObjectTagging: "keep=source"},
+				})
+				gr := migrationTestReader(t, z, 0, oi)
+				set := z.serverPools[0].getHashedSet(oi.Name)
+				original := set.getDisks
+				disks := append([]StorageAPI(nil), original()...)
+				for i := range disks {
+					disks[i] = consistencyReadFaultDisk{StorageAPI: disks[i], bucket: bucket, object: oi.Name}
+				}
+				set.getDisks = func() []StorageAPI { return disks }
+				defer func() { set.getDisks = original }()
+				err := move(t.Context(), 0, bucket, gr)
+				var quorum InsufficientReadQuorum
+				if !errors.As(err, &quorum) {
+					t.Fatalf("unreadable pool must fail migration with read quorum error: %v", err)
+				}
+				set.getDisks = original
+				migrationTestStored(t, z, 0, oi, "keep=source", "", false, "payload")
+				if _, err := z.serverPools[1].GetObjectInfo(t.Context(), bucket, oi.Name, ObjectOptions{VersionID: oi.VersionID}); !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+					t.Fatalf("failed migration committed a target version: %v", err)
+				}
+			})
+		}
+	}
+}
@@ -0,0 +1,721 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"context"
+	"crypto/md5"
+	"encoding/base64"
+	"fmt"
+	"maps"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	xhttp "github.com/minio/minio/internal/http"
+	"github.com/minio/minio/internal/kms"
+)
+
+// Change allocation capacity only; metadata and data use real fixture disks.
+// Restoring the adapters also lets encrypted fixtures move the write target
+// between requests without changing the production allocation policy.
+type conditionalPutCapacityDisk struct {
+	StorageAPI
+	full bool
+}
+
+func (d conditionalPutCapacityDisk) DiskInfo(ctx context.Context, opts DiskInfoOptions) (DiskInfo, error) {
+	info, err := d.StorageAPI.DiskInfo(ctx, opts)
+	info.Total, info.Used = 1<<40, 0
+	if d.full {
+		info.Used = info.Total - (1 << 20)
+	}
+	info.Free = info.Total - info.Used
+	return info, err
+}
+
+// Keep getDisks immutable while background IAM and storage readers use it.
+// GetDisks takes this same mutex when it copies the backing disk list.
+func conditionalPutSwapDisks(pool *erasureSets, object string, wrap func(StorageAPI) StorageAPI) func() {
+	setIndex := pool.getHashedSet(object).setIndex
+	pool.erasureDisksMu.Lock()
+	previous := pool.erasureDisks[setIndex]
+	disks := append([]StorageAPI(nil), previous...)
+	for i, disk := range disks {
+		disks[i] = wrap(disk)
+	}
+	pool.erasureDisks[setIndex] = disks
+	pool.erasureDisksMu.Unlock()
+	return func() {
+		pool.erasureDisksMu.Lock()
+		pool.erasureDisks[setIndex] = previous
+		pool.erasureDisksMu.Unlock()
+	}
+}
+
+func conditionalPutPool(t *testing.T, z *erasureServerPools, object string, target int) func() {
+	t.Helper()
+	var restore []func()
+	for i, pool := range z.serverPools {
+		restore = append(restore, conditionalPutSwapDisks(pool, object, func(disk StorageAPI) StorageAPI {
+			return conditionalPutCapacityDisk{StorageAPI: disk, full: i != target}
+		}))
+	}
+	return func() {
+		for _, fn := range restore {
+			fn()
+		}
+	}
+}
+
+func conditionalPutBucket(t *testing.T, z *erasureServerPools, mode string) (string, http.Handler) {
+	t.Helper()
+	bucket, router, err := initAPIHandlerTest(t.Context(), z, nil, MakeBucketOptions{VersioningEnabled: mode != "unversioned"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if mode == "suspended" {
+		if _, err := globalBucketMetadataSys.Update(t.Context(), bucket, bucketVersioningConfig,
+			[]byte(`<VersioningConfiguration><Status>Suspended</Status></VersioningConfiguration>`)); err != nil {
+			t.Fatal(err)
+		}
+	}
+	return bucket, router
+}
+
+func TestPoolsConditionalPutHTTP(t *testing.T) {
+	for _, mode := range []string{"unversioned", "versioned", "suspended"} {
+		t.Run(mode, func(t *testing.T) {
+			z, _ := consistencyPools(t)
+			bucket, router := conditionalPutBucket(t, z, mode)
+			for target := range 2 {
+				for _, tc := range []struct {
+					name                  string
+					oldCopy, localCurrent bool
+					condition             string
+					status                int
+				}{
+					{"stale-etag-accepted", true, false, "old", http.StatusPreconditionFailed},
+					{"current-etag-rejected", true, false, "current", http.StatusOK},
+					{"create-only-overwrites-other-pool", false, false, "none", http.StatusPreconditionFailed},
+					{"current-etag-missing-in-write-pool", false, false, "current", http.StatusOK},
+					{"control-current-in-write-pool", true, true, "current", http.StatusOK},
+					{"control-stale-etag-rejected", true, true, "old", http.StatusPreconditionFailed},
+					{"none-match-current-etag", true, false, "none-current", http.StatusPreconditionFailed},
+					{"none-match-old-etag", true, false, "none-old", http.StatusOK},
+				} {
+					t.Run(fmt.Sprintf("target=%d/%s", target, tc.name), func(t *testing.T) {
+						object := fmt.Sprintf("%d-%s", target, tc.name)
+						currentPool := 1 - target
+						if tc.localCurrent {
+							currentPool = target
+						}
+						opts := ObjectOptions{Versioned: mode == "versioned", VersionSuspended: mode == "suspended", MTime: UTCNow().Add(-time.Hour)}
+						oldETag := "absent-old"
+						if tc.oldCopy {
+							oldETag = putConsistencyObject(t, z, bucket, object, 1-currentPool, "old", opts).ETag
+						}
+						opts.MTime = UTCNow().Add(-time.Minute)
+						current := putConsistencyObject(t, z, bucket, object, currentPool, "current", opts)
+						defer conditionalPutPool(t, z, object, target)()
+						idx, err := z.getWritePoolIdx(t.Context(), bucket, object, 11, false)
+						if err != nil || idx != target {
+							t.Fatalf("allocation target=%d: idx=%d err=%v", target, idx, err)
+						}
+						headers := map[string]string{xhttp.IfMatch: fmt.Sprintf("%q", current.ETag)}
+						if tc.condition == "old" {
+							headers[xhttp.IfMatch] = fmt.Sprintf("%q", oldETag)
+						}
+						if tc.condition == "none" {
+							headers = map[string]string{xhttp.IfNoneMatch: "*"}
+						}
+						if tc.condition == "none-current" {
+							headers = map[string]string{xhttp.IfNoneMatch: fmt.Sprintf("%q", current.ETag)}
+						}
+						if tc.condition == "none-old" {
+							headers = map[string]string{xhttp.IfNoneMatch: fmt.Sprintf("%q", oldETag)}
+						}
+						url := getPutObjectURL("", bucket, object)
+						before := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+						if before.Code != http.StatusOK || before.Body.String() != "current" || multipartConditionResponseETag(before) != current.ETag {
+							t.Fatalf("invalid current object: %d %q %v", before.Code, before.Body.String(), before.Header())
+						}
+						put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", headers)
+						if put.Code != tc.status {
+							t.Errorf("PUT status=%d want=%d: %s", put.Code, tc.status, put.Body.String())
+						}
+						if put.Code == http.StatusPreconditionFailed {
+							multipartConditionError(t, put, "PreconditionFailed")
+							if multipartConditionResponseETag(put) != current.ETag || put.Header().Get(xhttp.LastModified) != before.Header().Get(xhttp.LastModified) {
+								t.Errorf("412 headers do not describe current object: %v", put.Header())
+							}
+						}
+						get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+						wantBody, wantETag := "current", current.ETag
+						if tc.status == http.StatusOK {
+							wantBody, wantETag = "replacement", fmt.Sprintf("%x", md5.Sum([]byte("replacement")))
+						}
+						t.Logf("PUT %d; GET %d bytes=%q ETag=%s", put.Code, get.Code, get.Body.String(), multipartConditionResponseETag(get))
+						if get.Code != http.StatusOK || get.Body.String() != wantBody || multipartConditionResponseETag(get) != wantETag {
+							t.Errorf("GET=%d bytes=%q ETag=%s; want %q %s", get.Code, get.Body.String(), multipartConditionResponseETag(get), wantBody, wantETag)
+						}
+					})
+				}
+			}
+		})
+	}
+}
+
+func TestPoolsConditionalPutHTTPAbsence(t *testing.T) {
+	for _, state := range []string{"missing", "uuid-marker", "null-marker"} {
+		for _, match := range []bool{false, true} {
+			t.Run(fmt.Sprintf("%s/match=%t", state, match), func(t *testing.T) {
+				z, _ := consistencyPools(t)
+				mode := "versioned"
+				if state == "null-marker" {
+					mode = "suspended"
+				}
+				bucket, router := conditionalPutBucket(t, z, mode)
+				object := "absent-key"
+				if state != "missing" {
+					putConsistencyObject(t, z, bucket, object, 0, "old", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Hour)})
+					vid := mustGetUUID()
+					if state == "null-marker" {
+						vid = nullVersionID
+					}
+					if _, err := z.serverPools[1].DeleteObject(t.Context(), bucket, object, ObjectOptions{Versioned: true, VersionID: vid, DeleteMarker: true, MTime: UTCNow().Add(-time.Minute)}); err != nil {
+						t.Fatal(err)
+					}
+				}
+				defer conditionalPutPool(t, z, object, 0)()
+				headers := map[string]string{xhttp.IfNoneMatch: "*"}
+				want := http.StatusOK
+				if match {
+					headers = map[string]string{xhttp.IfMatch: "*"}
+					want = http.StatusNotFound
+				}
+				url := getPutObjectURL("", bucket, object)
+				put := multipartConditionRequest(t, router, http.MethodPut, url, "new", headers)
+				if put.Code != want {
+					t.Fatalf("PUT %d want %d: %s", put.Code, want, put.Body.String())
+				}
+				get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+				if match {
+					multipartConditionError(t, put, "NoSuchKey")
+					if get.Code != http.StatusNotFound {
+						t.Fatalf("failed PUT exposed data: %d %s", get.Code, get.Body.String())
+					}
+				} else if get.Code != http.StatusOK || get.Body.String() != "new" || multipartConditionResponseETag(get) != multipartConditionResponseETag(put) {
+					t.Fatalf("successful create GET: %d %q %v", get.Code, get.Body.String(), get.Header())
+				}
+			})
+		}
+	}
+}
+
+func TestPoolsConditionalPutUnreadable(t *testing.T) {
+	for faultPool := range 2 {
+		for _, present := range []bool{false, true} {
+			for _, match := range []bool{false, true} {
+				t.Run(fmt.Sprintf("fault-pool=%d/present=%t/match=%t", faultPool, present, match), func(t *testing.T) {
+					z, _ := consistencyPools(t)
+					bucket, router := conditionalPutBucket(t, z, "unversioned")
+					object := "unreadable-key"
+					if present {
+						putConsistencyObject(t, z, bucket, object, 0, "old", ObjectOptions{MTime: UTCNow().Add(-time.Hour)})
+						putConsistencyObject(t, z, bucket, object, 1, "current", ObjectOptions{MTime: UTCNow().Add(-time.Minute)})
+					}
+					defer conditionalPutPool(t, z, object, 0)()
+					restoreFault := conditionalPutSwapDisks(z.serverPools[faultPool], object, func(disk StorageAPI) StorageAPI {
+						return consistencyReadFaultDisk{StorageAPI: disk, bucket: bucket, object: object}
+					})
+					defer restoreFault()
+					headers := map[string]string{xhttp.IfNoneMatch: "*"}
+					if match {
+						headers = map[string]string{xhttp.IfMatch: "*"}
+					}
+					url := getPutObjectURL("", bucket, object)
+					put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", headers)
+					if put.Code != http.StatusServiceUnavailable {
+						t.Errorf("unverified PUT must fail: %d %s", put.Code, put.Body.String())
+					}
+					called := 0
+					_, err := z.PutObject(t.Context(), bucket, object, mustGetPutObjReader(t, strings.NewReader("replacement"), 11, "", ""), ObjectOptions{HasIfMatch: match, CheckPrecondFn: func(ObjectInfo) bool { called++; return false }})
+					if !isErrReadQuorum(err) || called != 0 {
+						t.Errorf("lookup error=%v callback calls=%d", err, called)
+					}
+					restoreFault()
+					get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+					if present {
+						if get.Code != http.StatusOK || get.Body.String() != "current" || multipartConditionResponseETag(get) != fmt.Sprintf("%x", md5.Sum([]byte("current"))) {
+							t.Fatalf("failed PUT changed object: %d %q", get.Code, get.Body.String())
+						}
+					} else if get.Code != http.StatusNotFound {
+						t.Fatalf("failed PUT created object: %d %q", get.Code, get.Body.String())
+					}
+				})
+			}
+		}
+	}
+}
+
+func TestPoolsConditionalPutEncryptedETag(t *testing.T) {
+	for _, kind := range []string{"SSE-C", "SSE-S3", "SSE-KMS"} {
+		t.Run(kind, func(t *testing.T) {
+			z, _ := consistencyPools(t)
+			bucket, router := conditionalPutBucket(t, z, "unversioned")
+			oldKMS, oldTLS := GlobalKMS, globalIsTLS
+			GlobalKMS, globalIsTLS = kms.NewStub("conditional-put-key"), true
+			defer func() { GlobalKMS, globalIsTLS = oldKMS, oldTLS }()
+			headers := map[string]string{xhttp.AmzServerSideEncryption: xhttp.AmzEncryptionAES}
+			readHeaders := map[string]string{}
+			if kind == "SSE-C" {
+				key := bytes.Repeat([]byte{0x42}, 32)
+				digest := md5.Sum(key)
+				headers = map[string]string{
+					xhttp.AmzServerSideEncryptionCustomerAlgorithm: xhttp.AmzEncryptionAES,
+					xhttp.AmzServerSideEncryptionCustomerKey:       base64.StdEncoding.EncodeToString(key),
+					xhttp.AmzServerSideEncryptionCustomerKeyMD5:    base64.StdEncoding.EncodeToString(digest[:]),
+				}
+				readHeaders = maps.Clone(headers)
+			}
+			if kind == "SSE-KMS" {
+				headers[xhttp.AmzServerSideEncryption] = xhttp.AmzEncryptionKMS
+				headers[xhttp.AmzServerSideEncryptionKmsID] = "conditional-put-key"
+			}
+			object := "encrypted-key"
+			url := getPutObjectURL("", bucket, object)
+			restore := conditionalPutPool(t, z, object, 0)
+			old := multipartConditionRequest(t, router, http.MethodPut, url, "old", headers)
+			restore()
+			restore = conditionalPutPool(t, z, object, 1)
+			current := multipartConditionRequest(t, router, http.MethodPut, url, "current", headers)
+			restore()
+			if old.Code != http.StatusOK || current.Code != http.StatusOK {
+				t.Fatalf("encrypted setup: %d %s / %d %s", old.Code, old.Body.String(), current.Code, current.Body.String())
+			}
+			defer conditionalPutPool(t, z, object, 0)()
+			for _, stale := range []bool{true, false} {
+				h := maps.Clone(headers)
+				h[xhttp.IfMatch] = fmt.Sprintf("%q", multipartConditionResponseETag(current))
+				wantStatus, wantBody, wantETag := http.StatusOK, "replacement", ""
+				if stale {
+					h[xhttp.IfMatch] = fmt.Sprintf("%q", multipartConditionResponseETag(old))
+					wantStatus, wantBody, wantETag = http.StatusPreconditionFailed, "current", multipartConditionResponseETag(current)
+				}
+				put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", h)
+				if put.Code != wantStatus {
+					t.Fatalf("encrypted condition: %d want %d: %s", put.Code, wantStatus, put.Body.String())
+				}
+				if !stale {
+					wantETag = multipartConditionResponseETag(put)
+				}
+				get := multipartConditionRequest(t, router, http.MethodGet, url, "", readHeaders)
+				if get.Code != http.StatusOK || get.Body.String() != wantBody || multipartConditionResponseETag(get) != wantETag {
+					t.Fatalf("encrypted GET: %d %q %v", get.Code, get.Body.String(), get.Header())
+				}
+			}
+		})
+	}
+}
+
+func TestPoolsConditionalPutVersionSelection(t *testing.T) {
+	for _, kind := range []string{"public-version", "replica", "replica-preserve-etag", "movement", "no-lock", "tie", "draining"} {
+		t.Run(kind, func(t *testing.T) {
+			z, bucket := consistencyPools(t)
+			object := "version-selection"
+			addressed := putConsistencyObject(t, z, bucket, object, 1, "addressed", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Hour)})
+			current := putConsistencyObject(t, z, bucket, object, 1, "current", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Minute)})
+			opts := ObjectOptions{Versioned: true, VersionID: addressed.VersionID, HasIfMatch: true}
+			want := current
+			switch kind {
+			case "replica":
+				opts.ReplicaLockReconcile, opts.ReplicationRequest = true, true
+				want = addressed
+			case "replica-preserve-etag":
+				opts.PreserveETag = addressed.ETag
+				opts.ReplicaLockReconcile, opts.ReplicationRequest = true, true
+				want = addressed
+			case "movement":
+				opts.DataMovement, opts.SrcPoolIdx = true, 1
+				want = addressed
+			case "tie":
+				want = putConsistencyObject(t, z, bucket, object, 0, "tie-winner", ObjectOptions{Versioned: true, MTime: current.ModTime})
+			case "draining":
+				z.poolMetaMutex.Lock()
+				z.poolMeta.Pools[1].Decommission = &PoolDecommissionInfo{}
+				z.poolMetaMutex.Unlock()
+			}
+			defer conditionalPutPool(t, z, object, 0)()
+			ctx := t.Context()
+			if kind == "no-lock" {
+				lk := z.NewNSLock(bucket, object)
+				lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+				if err != nil {
+					t.Fatal(err)
+				}
+				defer lk.Unlock(lkctx)
+				ctx, opts.NoLock = lkctx.Context(), true
+			}
+			called := 0
+			opts.UserDefined = make(map[string]string)
+			opts.CheckPrecondFn = func(oi ObjectInfo) bool {
+				called++
+				if oi.ETag != want.ETag || oi.VersionID != want.VersionID {
+					t.Errorf("comparison ETag/version=%s/%s want %s/%s", oi.ETag, oi.VersionID, want.ETag, want.VersionID)
+				}
+				return oi.ETag != want.ETag
+			}
+			oi, err := z.PutObject(ctx, bucket, object, mustGetPutObjReader(t, strings.NewReader("replacement"), 11, "", ""), opts)
+			if err != nil || called != 1 {
+				t.Fatalf("PUT err=%v callback calls=%d", err, called)
+			}
+			if oi.VersionID != addressed.VersionID {
+				t.Fatalf("destination version changed: %s", oi.VersionID)
+			}
+			if opts.PreserveETag != "" && oi.ETag != opts.PreserveETag {
+				t.Fatalf("PreserveETag changed: %s", oi.ETag)
+			}
+		})
+	}
+}
+
+func TestPoolsConditionalPutReplicaDuplicateHTTP(t *testing.T) {
+	for _, null := range []bool{false, true} {
+		t.Run(fmt.Sprintf("null=%t", null), func(t *testing.T) {
+			z, _ := consistencyPools(t)
+			bucket, router := conditionalPutBucket(t, z, "versioned")
+			object := "replica-duplicate"
+			vid := mustGetUUID()
+			if null {
+				vid = nullVersionID
+			}
+			addressed := putConsistencyObject(t, z, bucket, object, 1, "addressed", ObjectOptions{Versioned: true, VersionID: vid, MTime: UTCNow().Add(-time.Hour)})
+			current := putConsistencyObject(t, z, bucket, object, 1, "current", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Minute)})
+			defer conditionalPutPool(t, z, object, 0)()
+			headers := map[string]string{
+				xhttp.MinIOSourceReplicationRequest: "true",
+				xhttp.AmzBucketReplicationStatus:    "REPLICA",
+				xhttp.MinIOSourceETag:               addressed.ETag,
+				xhttp.MinIOSourceMTime:              addressed.ModTime.Format(time.RFC3339Nano),
+			}
+			url := getPutObjectURL("", bucket, object)
+			put := multipartConditionRequest(t, router, http.MethodPut, url+"?versionId="+vid, "addressed", headers)
+			if put.Code != http.StatusPreconditionFailed {
+				t.Fatalf("replica duplicate: %d %s", put.Code, put.Body.String())
+			}
+			multipartConditionError(t, put, "PreconditionFailed")
+			get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+			if get.Code != http.StatusOK || get.Body.String() != "current" || multipartConditionResponseETag(get) != current.ETag {
+				t.Fatalf("duplicate changed current object: %d %q", get.Code, get.Body.String())
+			}
+			get = multipartConditionRequest(t, router, http.MethodGet, url+"?versionId="+vid, "", nil)
+			if get.Code != http.StatusOK || get.Body.String() != "addressed" || multipartConditionResponseETag(get) != addressed.ETag {
+				t.Fatalf("duplicate changed addressed version: %d %q", get.Code, get.Body.String())
+			}
+		})
+	}
+}
+
+func TestPoolsConditionalPutConcurrentHTTP(t *testing.T) {
+	z, _ := consistencyPools(t)
+	bucket, router := conditionalPutBucket(t, z, "unversioned")
+	for _, match := range []bool{false, true} {
+		for iteration := range 5 {
+			t.Run(fmt.Sprintf("match=%t/iteration=%d", match, iteration), func(t *testing.T) {
+				object := fmt.Sprintf("concurrent-%t-%d", match, iteration)
+				headers := map[string]string{xhttp.IfNoneMatch: "*"}
+				if match {
+					oi := putConsistencyObject(t, z, bucket, object, 1, "old", ObjectOptions{MTime: UTCNow().Add(-time.Minute)})
+					headers = map[string]string{xhttp.IfMatch: fmt.Sprintf("%q", oi.ETag)}
+				}
+				defer conditionalPutPool(t, z, object, 0)()
+				start := make(chan struct{})
+				results := make(chan *httptest.ResponseRecorder, 2)
+				url := getPutObjectURL("", bucket, object)
+				for i := range 2 {
+					body := fmt.Sprintf("writer-%d", i)
+					req, err := newTestSignedRequestV4(http.MethodPut, url, int64(len(body)), strings.NewReader(body), globalActiveCred.AccessKey, globalActiveCred.SecretKey, headers)
+					if err != nil {
+						t.Fatal(err)
+					}
+					go func() {
+						<-start
+						rec := httptest.NewRecorder()
+						router.ServeHTTP(rec, req)
+						results <- rec
+					}()
+				}
+				close(start)
+				success, failed, winnerETag := 0, 0, ""
+				for range 2 {
+					result := <-results
+					switch result.Code {
+					case http.StatusOK:
+						success++
+						winnerETag = multipartConditionResponseETag(result)
+					case http.StatusPreconditionFailed:
+						failed++
+						multipartConditionError(t, result, "PreconditionFailed")
+					default:
+						t.Errorf("unexpected PUT %d: %s", result.Code, result.Body.String())
+					}
+				}
+				if success != 1 || failed != 1 {
+					t.Fatalf("success=%d precondition failures=%d", success, failed)
+				}
+				get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+				if get.Code != http.StatusOK || multipartConditionResponseETag(get) != winnerETag || fmt.Sprintf("%x", md5.Sum(get.Body.Bytes())) != winnerETag {
+					t.Fatalf("winner lost: %d %q %v", get.Code, get.Body.String(), get.Header())
+				}
+			})
+		}
+	}
+}
+
+func TestPoolsConditionalPutSerializesMutation(t *testing.T) {
+	for _, deletion := range []bool{false, true} {
+		t.Run(fmt.Sprintf("delete=%t", deletion), func(t *testing.T) {
+			z, bucket := consistencyPools(t)
+			object := "conditional-mutation"
+			current := putConsistencyObject(t, z, bucket, object, 1, "current", ObjectOptions{MTime: UTCNow().Add(-time.Minute)})
+			defer conditionalPutPool(t, z, object, 0)()
+			ctx, cancel := context.WithTimeout(t.Context(), 10*time.Second)
+			defer cancel()
+			gate := &consistencyGateReader{Reader: strings.NewReader("replacement"), entered: make(chan struct{}), resume: make(chan struct{})}
+			release := func() { gate.release.Do(func() { close(gate.resume) }) }
+			defer release()
+			reader := mustGetPutObjReader(t, gate, 11, "", "")
+			written := make(chan error, 1)
+			go func() {
+				_, err := z.PutObject(ctx, bucket, object, reader, ObjectOptions{HasIfMatch: true, CheckPrecondFn: func(oi ObjectInfo) bool { return oi.ETag != current.ETag }})
+				written <- err
+			}()
+			select {
+			case <-gate.entered:
+			case err := <-written:
+				t.Fatalf("PUT failed before body read: %v", err)
+			case <-ctx.Done():
+				t.Fatal(ctx.Err())
+			}
+			mutated := make(chan error, 1)
+			go func() {
+				var err error
+				if deletion {
+					_, err = z.DeleteObject(ctx, bucket, object, ObjectOptions{})
+				} else {
+					_, err = z.PutObjectMetadata(ctx, bucket, object, ObjectOptions{EvalMetadataFn: func(oi *ObjectInfo, _ error) (ReplicateDecision, error) {
+						oi.UserDefined["x-amz-meta-after-put"] = "present"
+						return ReplicateDecision{}, nil
+					}})
+				}
+				mutated <- err
+			}()
+			select {
+			case err := <-mutated:
+				t.Fatalf("mutation escaped PUT lock: %v", err)
+			case <-time.After(100 * time.Millisecond):
+			}
+			release()
+			if err := <-written; err != nil {
+				t.Fatal(err)
+			}
+			if err := <-mutated; err != nil {
+				t.Fatal(err)
+			}
+			oi, err := z.GetObjectInfo(ctx, bucket, object, ObjectOptions{})
+			if deletion {
+				if !isErrObjectNotFound(err) {
+					t.Fatalf("delete lost: %v", err)
+				}
+			} else if err != nil || oi.UserDefined["x-amz-meta-after-put"] != "present" || oi.ETag != fmt.Sprintf("%x", md5.Sum([]byte("replacement"))) {
+				t.Fatalf("metadata/PUT lost: %+v %v", oi, err)
+			}
+		})
+	}
+}
+
+func TestSinglePoolConditionalPutHTTP(t *testing.T) {
+	ctx, cancel := context.WithCancel(t.Context())
+	obj, dirs, err := prepareErasure16(ctx)
+	if err != nil {
+		cancel()
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	t.Cleanup(func() { cancel(); z.Shutdown(context.Background()); removeRoots(dirs) })
+	if !z.SinglePool() {
+		t.Fatal("fixture is not a single pool")
+	}
+	bucket, router := conditionalPutBucket(t, z, "unversioned")
+	object := "single-pool-condition"
+	defer conditionalPutPool(t, z, object, 0)()
+	url := getPutObjectURL("", bucket, object)
+	for _, tc := range []struct {
+		body, match, none string
+		status            int
+	}{
+		{"missing", "*", "", http.StatusNotFound},
+		{"first", "", "*", http.StatusOK},
+		{"blocked", "", "*", http.StatusPreconditionFailed},
+		{"blocked", "stale", "", http.StatusPreconditionFailed},
+		{"second", fmt.Sprintf("%x", md5.Sum([]byte("first"))), "", http.StatusOK},
+	} {
+		rec := multipartConditionRequest(t, router, http.MethodPut, url, tc.body, map[string]string{xhttp.IfMatch: tc.match, xhttp.IfNoneMatch: tc.none})
+		if rec.Code != tc.status {
+			t.Fatalf("PUT %d want %d: %s", rec.Code, tc.status, rec.Body.String())
+		}
+	}
+	get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+	if get.Code != http.StatusOK || get.Body.String() != "second" {
+		t.Fatalf("single-pool GET: %d %q", get.Code, get.Body.String())
+	}
+}
+
+// An internal replica callback without an addressed version is not a public
+// condition. Preserve its availability when another pool is unreadable. An
+// addressed replica already requires all pools for lock/tag reconciliation.
+func TestPoolsConditionalPutReplicaAvailability(t *testing.T) {
+	for _, addressed := range []bool{false, true} {
+		t.Run(fmt.Sprintf("addressed=%t", addressed), func(t *testing.T) {
+			z, _ := consistencyPools(t)
+			mode := "unversioned"
+			if addressed {
+				mode = "versioned"
+			}
+			bucket, router := conditionalPutBucket(t, z, mode)
+			object := "replica-availability"
+			oi := putConsistencyObject(t, z, bucket, object, 0, "old", ObjectOptions{Versioned: addressed, MTime: UTCNow().Add(-time.Minute)})
+			defer conditionalPutPool(t, z, object, 0)()
+			restoreFault := conditionalPutSwapDisks(z.serverPools[1], object, func(disk StorageAPI) StorageAPI {
+				return consistencyReadFaultDisk{StorageAPI: disk, bucket: bucket, object: object}
+			})
+			defer restoreFault()
+			headers := map[string]string{
+				xhttp.MinIOSourceReplicationRequest: "true",
+				xhttp.AmzBucketReplicationStatus:    "REPLICA",
+				xhttp.MinIOSourceETag:               fmt.Sprintf("%x", md5.Sum([]byte("replacement"))),
+			}
+			url := getPutObjectURL("", bucket, object)
+			wantStatus, wantBody := http.StatusOK, "replacement"
+			if addressed {
+				url += "?versionId=" + oi.VersionID
+				wantStatus, wantBody = http.StatusServiceUnavailable, "old"
+			}
+			put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", headers)
+			if put.Code != wantStatus {
+				t.Fatalf("replica PUT: %d want %d: %s", put.Code, wantStatus, put.Body.String())
+			}
+			restoreFault()
+			get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+			if get.Code != http.StatusOK || get.Body.String() != wantBody {
+				t.Fatalf("replica GET: %d %q", get.Code, get.Body.String())
+			}
+		})
+	}
+}
+
+func TestPoolsConditionalPutDestinationVersionHTTP(t *testing.T) {
+	z, _ := consistencyPools(t)
+	bucket, router := conditionalPutBucket(t, z, "versioned")
+	object := "client-version"
+	old := putConsistencyObject(t, z, bucket, object, 0, "old", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Hour)})
+	current := putConsistencyObject(t, z, bucket, object, 1, "current", ObjectOptions{Versioned: true, MTime: UTCNow().Add(-time.Minute)})
+	defer conditionalPutPool(t, z, object, 0)()
+	url := getPutObjectURL("", bucket, object) + "?versionId=" + old.VersionID
+	for _, stale := range []bool{true, false} {
+		tag, status := current.ETag, http.StatusOK
+		if stale {
+			tag, status = old.ETag, http.StatusPreconditionFailed
+		}
+		put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", map[string]string{xhttp.IfMatch: fmt.Sprintf("%q", tag)})
+		if put.Code != status {
+			t.Fatalf("version-addressed public PUT: %d want %d: %s", put.Code, status, put.Body.String())
+		}
+		if got := put.Header()[xhttp.AmzVersionID]; !stale && (len(got) != 1 || got[0] != old.VersionID) {
+			t.Fatalf("write version changed: %v", put.Header())
+		}
+	}
+	get := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+	if get.Code != http.StatusOK || get.Body.String() != "replacement" {
+		t.Fatalf("addressed GET: %d %q", get.Code, get.Body.String())
+	}
+}
+
+func TestPoolsConditionalPutDeleteMarkerTie(t *testing.T) {
+	for markerPool := range 2 {
+		t.Run(fmt.Sprintf("marker-pool=%d", markerPool), func(t *testing.T) {
+			z, _ := consistencyPools(t)
+			bucket, router := conditionalPutBucket(t, z, "versioned")
+			object := "marker-tie"
+			mtime := UTCNow().Add(-time.Minute)
+			putConsistencyObject(t, z, bucket, object, 1-markerPool, "live", ObjectOptions{Versioned: true, MTime: mtime})
+			if _, err := z.serverPools[markerPool].DeleteObject(t.Context(), bucket, object, ObjectOptions{Versioned: true, VersionID: mustGetUUID(), DeleteMarker: true, MTime: mtime}); err != nil {
+				t.Fatal(err)
+			}
+			defer conditionalPutPool(t, z, object, 0)()
+			url := getPutObjectURL("", bucket, object)
+			before := multipartConditionRequest(t, router, http.MethodGet, url, "", nil)
+			want := http.StatusPreconditionFailed
+			if markerPool == 0 {
+				want = http.StatusOK
+				if before.Code != http.StatusNotFound {
+					t.Fatalf("GET tie: %d", before.Code)
+				}
+			} else if before.Code != http.StatusOK {
+				t.Fatalf("GET tie: %d", before.Code)
+			}
+			put := multipartConditionRequest(t, router, http.MethodPut, url, "replacement", map[string]string{xhttp.IfNoneMatch: "*"})
+			if put.Code != want {
+				t.Fatalf("PUT tie: %d want %d: %s", put.Code, want, put.Body.String())
+			}
+		})
+	}
+}
+
+// Overlap fixture changes with the real IAM Walk reader instead of relying on
+// its periodic refresh timer to expose an unsynchronized disk-adapter swap.
+func TestPoolsConditionalPutFixtureConcurrentIAM(t *testing.T) {
+	z, _ := consistencyPools(t)
+	conditionalPutBucket(t, z, "unversioned")
+	ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+	defer cancel()
+	started, finished := make(chan struct{}), make(chan error, 1)
+	iam := globalIAMSys
+	go func() {
+		close(started)
+		for range 20 {
+			if err := iam.Load(ctx, false); err != nil {
+				finished <- err
+				return
+			}
+		}
+		finished <- nil
+	}()
+	<-started
+	for i := range 5000 {
+		restore := conditionalPutPool(t, z, "fixture-concurrent-iam", i%2)
+		restore()
+	}
+	if err := <-finished; err != nil {
+		t.Fatal(err)
+	}
+}
@@ -865,7 +865,7 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, poolIdx int, b
 	if oi.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, oi.Name, ObjectOptions{
 			VersionID:    oi.VersionID,
-			UserDefined:  oi.UserDefined,
+			UserDefined:  migrationObjectMetadata(oi),
 			NoAuditLog:   true,
 			DataMovement: true,
 			SrcPoolIdx:   poolIdx,
@@ -924,7 +924,7 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, poolIdx int, b
 			DataMovement: true,
 			VersionID:    oi.VersionID,
 			MTime:        oi.ModTime,
-			UserDefined:  oi.UserDefined,
+			UserDefined:  migrationObjectMetadata(oi),
 			PreserveETag: oi.ETag, // Preserve original ETag to ensure same metadata.
 			IndexCB: func() []byte {
 				return oi.Parts[0].Index // Preserve part Index to ensure decompression works.
@@ -1149,6 +1149,40 @@ func (z *erasureServerPools) PutObject(ctx context.Context, bucket string, objec
 	}
 	opts.NoLock = true

+	// Public write conditions compare the logical current object while the
+	// pools-layer write lock is held. The destination selected by capacity may
+	// be empty or stale, and draining pools can still hold the current object.
+	// Replica callbacks retain their existing addressed-version semantics and
+	// metadata reconciliation at the set layer.
+	if opts.CheckPrecondFn != nil && !opts.ReplicationRequest &&
+		!opts.ReplicaLockReconcile && !opts.DataMovement {
+		copies, lerr := z.objectPoolInfos(ctx, bucket, object, ObjectOptions{
+			VersionID:        "", // Compare the current object, not the write's version.
+			Versioned:        opts.Versioned,
+			VersionSuspended: opts.VersionSuspended,
+			NoAuditLog:       true,
+		})
+		var latest ObjectInfo
+		if lerr == nil {
+			latest = copies[0].ObjInfo
+			if latest.DeleteMarker {
+				lerr = toObjectErr(errFileNotFound, bucket, object)
+			}
+		}
+		// An unreadable pool may hold the newest object; it is not absence.
+		if lerr != nil && !isErrObjectNotFound(lerr) && !isErrVersionNotFound(lerr) {
+			return ObjectInfo{}, lerr
+		}
+		if lerr == nil && opts.CheckPrecondFn(latest) {
+			return ObjectInfo{}, PreConditionFailed{}
+		}
+		if lerr != nil && opts.HasIfMatch {
+			return ObjectInfo{}, lerr
+		}
+		// Do not repeat an accepted condition against the destination's copy.
+		opts.CheckPrecondFn = nil
+	}
+
 	idx, err := z.getWritePoolIdx(ctx, bucket, object, data.Size(), true)
 	if err != nil {
 		return ObjectInfo{}, err
@@ -1205,9 +1239,12 @@ func (z *erasureServerPools) DeleteObject(ctx context.Context, bucket string, ob
 		return ObjectInfo{}, z.deletePrefix(ctx, bucket, object)
 	}

-	// Reconcile ordinary addressed-version deletes independently of pool movement.
+	// Resolve a physical purge by its addressed version even on a replica
+	// receiver: latest-key routing can select a different pool or miss copies.
+	// Marker creation and specialized movement/scanner operations retain their
+	// existing routing.
 	reconcileVersion := opts.VersionID != "" && !opts.DataMovement &&
-		!opts.ReplicationRequest && !opts.Expiration.Expire && !opts.InclFreeVersions
+		(!opts.ReplicationRequest || opts.isVersionPurge()) && !opts.Expiration.Expire && !opts.InclFreeVersions
 	if !z.SinglePool() && (opts.CheckPrecondFn != nil || reconcileVersion) {
 		return z.deleteObjectReconciled(ctx, bucket, object, opts)
 	}
@@ -1220,6 +1257,13 @@ func (z *erasureServerPools) DeleteObject(ctx context.Context, bucket string, ob
 		if _, ok := err.(InsufficientReadQuorum); ok {
 			return objInfo, InsufficientWriteQuorum{}
 		}
+		// Lookup can return before the set's purge confirmation. Check here,
+		// before any callback can change the request into a metadata update.
+		if opts.isVersionPurge() && (isErrObjectNotFound(err) || isErrVersionNotFound(err)) {
+			if quorumErr := z.checkPurgeAbsent(ctx, bucket, object, opts.VersionID); quorumErr != nil {
+				return objInfo, quorumErr
+			}
+		}
 		// A conditional (If-Match) delete addressing a specific version treats an
 		// absent key as an absent version. getPoolInfoExistingWithOpts strips
 		// VersionID, so a missing key surfaces ObjectNotFound here even for a
@@ -1250,6 +1294,11 @@ func (z *erasureServerPools) DeleteObject(ctx context.Context, bucket string, ob
 			if verr != nil && (!isErrMethodNotAllowed(verr) || !vi.DeleteMarker) {
 				// Genuine read failure for the addressed version: a missing
 				// version -> VersionNotFound (NoSuchVersion), read-quorum loss, etc.
+				if opts.isVersionPurge() && (isErrObjectNotFound(verr) || isErrVersionNotFound(verr)) {
+					if quorumErr := z.checkPurgeAbsent(ctx, bucket, object, opts.VersionID); quorumErr != nil {
+						return objInfo, quorumErr
+					}
+				}
 				return objInfo, verr
 			}
 			// verr is nil for a live version, or MethodNotAllowed with a populated
@@ -1857,7 +1906,41 @@ func (z *erasureServerPools) ListMultipartUploads(ctx context.Context, bucket, p
 	if err := checkListMultipartArgs(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter); err != nil {
 		return ListMultipartsInfo{}, err
 	}
+	if _, err := z.GetBucketInfo(ctx, bucket, BucketOptions{}); err != nil {
+		return ListMultipartsInfo{}, toObjectErr(err, bucket)
+	}
+	if globalAPIConfig.getMultipartListingLegacy() {
+		return z.listMultipartUploadsLegacy(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
+	}
+	scan, err := startMultipartScan(ctx, false)
+	if err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	defer scan.close()

+	var uploads []MultipartInfo
+	var keyless bool
+	for idx, pool := range z.serverPools {
+		if z.IsSuspended(idx) {
+			continue
+		}
+		poolUploads, poolKeyless, err := pool.scanMultipartUploads(scan, bucket, idx)
+		if err != nil {
+			return ListMultipartsInfo{}, err
+		}
+		uploads = append(uploads, poolUploads...)
+		keyless = keyless || poolKeyless
+	}
+
+	// The old format cannot be enumerated authoritatively. Migration mode is
+	// explicit: another bucket must never silently change this API's semantics.
+	if keyless {
+		return ListMultipartsInfo{}, errMultipartListingLegacy
+	}
+	return paginateMultipartUploads(uploads, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads), nil
+}
+
+func (z *erasureServerPools) listMultipartUploadsLegacy(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (ListMultipartsInfo, error) {
 	poolResult := ListMultipartsInfo{}
 	poolResult.MaxUploads = maxUploads
 	poolResult.KeyMarker = keyMarker
@@ -1883,15 +1966,14 @@ func (z *erasureServerPools) ListMultipartUploads(ctx context.Context, bucket, p
 	}

 	if z.SinglePool() {
-		return z.serverPools[0].ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
+		return z.serverPools[0].getHashedSet(prefix).listMultipartUploadsExact(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
 	}

 	for idx, pool := range z.serverPools {
 		if z.IsSuspended(idx) {
 			continue
 		}
-		result, err := pool.ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker,
-			delimiter, maxUploads)
+		result, err := pool.getHashedSet(prefix).listMultipartUploadsExact(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
 		if err != nil {
 			return result, err
 		}
@@ -1927,7 +2009,7 @@ func (z *erasureServerPools) NewMultipartUpload(ctx context.Context, bucket, obj
 			continue
 		}

-		result, err := pool.ListMultipartUploads(ctx, bucket, object, "", "", "", maxUploadsList)
+		result, err := pool.listMultipartUploadsExact(ctx, bucket, object)
 		if err != nil {
 			return nil, err
 		}
@@ -2089,13 +2171,18 @@ func (z *erasureServerPools) AbortMultipartUpload(ctx context.Context, bucket, o
 	if err := checkAbortMultipartArgs(ctx, bucket, object, uploadID); err != nil {
 		return err
 	}
+	if _, err := z.GetBucketInfo(ctx, bucket, BucketOptions{}); err != nil {
+		return toObjectErr(err, bucket)
+	}

-	defer func() {
+	// Unlock cancels the derived lock context before this notification runs.
+	// Keep the request context so successful cancellation reaches peer caches.
+	defer func(ctx context.Context) {
 		if err == nil {
 			z.mpCache.Delete(uploadID)
 			globalNotificationSys.DeleteUploadID(ctx, uploadID)
 		}
-	}()
+	}(ctx)

 	lk := z.NewNSLock(bucket, pathJoin(object, uploadID))
 	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
@@ -2105,23 +2192,28 @@ func (z *erasureServerPools) AbortMultipartUpload(ctx context.Context, bucket, o
 	ctx = lkctx.Context()
 	defer lk.Unlock(lkctx)

-	if z.SinglePool() {
-		return z.serverPools[0].AbortMultipartUpload(ctx, bucket, object, uploadID, opts)
-	}
-
+	legacy := globalAPIConfig.getMultipartListingLegacy()
+	found := false
+	var firstErr error
 	for idx, pool := range z.serverPools {
 		if z.IsSuspended(idx) {
 			continue
 		}
-		err := pool.AbortMultipartUpload(ctx, bucket, object, uploadID, opts)
-		if err == nil {
-			return nil
+		poolFound, err := pool.getHashedSet(object).abortMultipartUpload(ctx, bucket, object, uploadID, opts, legacy)
+		if legacy && (poolFound || err != nil) {
+			// Match the released first-matching-pool behavior.
+			return err
 		}
-		if _, ok := err.(InvalidUploadID); ok {
-			// upload id not found move to next pool
-			continue
+		found = found || poolFound
+		if err != nil && firstErr == nil {
+			firstErr = err
 		}
-		return err
+	}
+	if firstErr != nil {
+		return firstErr
+	}
+	if found {
+		return nil
 	}
 	return InvalidUploadID{
 		Bucket:   bucket,
@@ -880,10 +880,40 @@ func (s *erasureSets) CopyObject(ctx context.Context, srcBucket, srcObject, dstB
 }

 func (s *erasureSets) ListMultipartUploads(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
-	// In list multipart uploads we are going to treat input prefix as the object,
-	// this means that we are not supporting directory navigation.
-	set := s.getHashedSet(prefix)
-	return set.ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
+	if err := checkListMultipartArgs(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter); err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	scan, err := startMultipartScan(ctx, false)
+	if err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	defer scan.close()
+	uploads, legacy, err := s.scanMultipartUploads(scan, bucket, 0)
+	if err != nil {
+		return ListMultipartsInfo{}, err
+	}
+	if legacy {
+		return ListMultipartsInfo{}, errMultipartListingLegacy
+	}
+	return paginateMultipartUploads(uploads, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads), nil
+}
+
+func (s *erasureSets) scanMultipartUploads(scan *multipartScan, bucket string, poolIdx int) ([]MultipartInfo, bool, error) {
+	var uploads []MultipartInfo
+	var keyless bool
+	for i, set := range s.sets {
+		setUploads, setKeyless, err := set.scanMultipartUploads(scan, bucket, poolIdx, i)
+		if err != nil {
+			return nil, false, err
+		}
+		uploads = append(uploads, setUploads...)
+		keyless = keyless || setKeyless
+	}
+	return uploads, keyless, nil
+}
+
+func (s *erasureSets) listMultipartUploadsExact(ctx context.Context, bucket, object string) (ListMultipartsInfo, error) {
+	return s.getHashedSet(object).listMultipartUploadsExact(ctx, bucket, object, "", "", "", maxUploadsList)
 }

 // Initiate a new multipart upload on a hashedSet based on object name.
@@ -0,0 +1,87 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+
+	"github.com/google/uuid"
+	"github.com/minio/minio/internal/bucket/replication"
+)
+
+// isVersionPurge distinguishes physical removal from marker creation and
+// replication-state updates. Call it again after metadata callbacks, which
+// can change an ordinary deletion into a pending purge update.
+func (o ObjectOptions) isVersionPurge() bool {
+	if o.VersionID == "" || o.VersionID == nullVersionID || o.DeleteMarker ||
+		o.DeletePrefix || o.DataMovement || o.InclFreeVersions || o.Expiration.Expire ||
+		o.Transition != (TransitionOptions{}) {
+		return false
+	}
+	id, err := uuid.Parse(o.VersionID)
+	if err != nil || id == uuid.Nil {
+		return false
+	}
+	purge := o.VersionPurgeStatus()
+	if purge == replication.VersionPurgeComplete {
+		return true
+	}
+	if !purge.Empty() || o.DeleteReplication.VersionPurgeStatusInternal != "" {
+		return false
+	}
+	status := o.DeleteMarkerReplicationStatus()
+	return (status.Empty() && o.DeleteReplication.ReplicationStatusInternal == "") ||
+		(status == replication.Replica && o.ReplicationRequest)
+}
+
+// confirmVersionAbsent is a read-only proof for an already-missing purge.
+// Read quorum (or a synthesized NotFound) cannot acknowledge a write. Do not
+// delete here: unreadable minority copies may carry retention we cannot check.
+// The caller holds the object lock. This function never heals or enqueues work.
+func (er erasureObjects) confirmVersionAbsent(ctx context.Context, bucket, object, versionID string) (allAbsent bool, err error) {
+	disks := er.getDisks()
+	_, errs := readAllFileInfo(ctx, disks, "", bucket, object, versionID, false, false)
+	absent := 0
+	for _, err := range errs {
+		if err == errFileNotFound || err == errFileVersionNotFound {
+			absent++
+		}
+	}
+	if absent < len(disks)/2+1 {
+		return false, InsufficientWriteQuorum{}
+	}
+	return absent == len(disks), nil
+}
+
+// checkPurgeAbsent schedules recovery explicitly, outside the read-only proof.
+func (er erasureObjects) checkPurgeAbsent(ctx context.Context, bucket, object, versionID string) error {
+	allAbsent, err := er.confirmVersionAbsent(ctx, bucket, object, versionID)
+	if !allAbsent {
+		er.addPartial(bucket, object, versionID)
+	}
+	return err
+}
+
+func (z *erasureServerPools) checkPurgeAbsent(ctx context.Context, bucket, object, versionID string) error {
+	for _, pool := range z.serverPools {
+		if err := pool.getHashedSet(object).checkPurgeAbsent(ctx, bucket, object, versionID); err != nil {
+			return err
+		}
+	}
+	return nil
+}
@@ -50,6 +50,7 @@ type apiConfig struct {
 	transitionWorkers      int

 	staleUploadsExpiry          time.Duration
+	multipartListingStrict      bool
 	staleUploadsCleanupInterval time.Duration
 	deleteCleanupInterval       time.Duration
 	enableODirect               bool
@@ -181,6 +182,7 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int, legacy bool) {
 	t.transitionWorkers = cfg.TransitionWorkers

 	t.staleUploadsExpiry = cfg.StaleUploadsExpiry
+	t.multipartListingStrict = cfg.MultipartListing == "strict"
 	t.deleteCleanupInterval = cfg.DeleteCleanupInterval
 	t.enableODirect = cfg.EnableODirect
 	t.gzipObjects = cfg.GzipObjects
@@ -206,6 +208,12 @@ func (t *apiConfig) odirectEnabled() bool {
 	return t.enableODirect
 }

+func (t *apiConfig) getMultipartListingLegacy() bool {
+	t.mu.RLock()
+	defer t.mu.RUnlock()
+	return !t.multipartListingStrict
+}
+
 func (t *apiConfig) shouldGzipObjects() bool {
 	t.mu.RLock()
 	defer t.mu.RUnlock()
@@ -0,0 +1,292 @@
+// Copyright (c) 2026 mr javad seydi
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"slices"
+	"testing"
+	"time"
+)
+
+func multipartUploadKeys(uploads []MultipartInfo) []string {
+	keys := make([]string, len(uploads))
+	for i := range uploads {
+		keys[i] = uploads[i].Object
+	}
+	return keys
+}
+
+func requireMultipartUploadKeys(t *testing.T, got ListMultipartsInfo, want ...string) {
+	t.Helper()
+	if keys := multipartUploadKeys(got.Uploads); !slices.Equal(keys, want) {
+		t.Fatalf("uploads = %v, want %v", keys, want)
+	}
+}
+
+func TestListMultipartUploadsS3Compatibility(t *testing.T) {
+	obj, dirs, err := prepareErasureSets32(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	setMultipartListingTestMode(t, false)
+	t.Cleanup(func() {
+		z.Shutdown(t.Context())
+		removeRoots(dirs)
+	})
+
+	const bucket = "multipart-list-compat"
+	if err = z.MakeBucket(t.Context(), bucket, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+
+	objects := []string{"t/a_b/p1", "t/a_b/p2", "t/c_d/p1", "u/x"}
+	sets := z.serverPools[0]
+	firstSet := sets.getHashedSetIndex(objects[0])
+	if !slices.ContainsFunc(objects[1:], func(object string) bool {
+		return sets.getHashedSetIndex(object) != firstSet
+	}) {
+		for n := 0; ; n++ {
+			object := fmt.Sprintf("v/cross-set-%d", n)
+			if sets.getHashedSetIndex(object) != firstSet {
+				objects = append(objects, object)
+				break
+			}
+		}
+	}
+
+	uploadIDs := make(map[string]string, len(objects))
+	for _, object := range objects {
+		mp, err := z.NewMultipartUpload(t.Context(), bucket, object, ObjectOptions{})
+		if err != nil {
+			t.Fatalf("NewMultipartUpload(%q): %v", object, err)
+		}
+		uploadIDs[object] = mp.UploadID
+	}
+
+	// Durable multipart metadata, rather than this node-local cache, must be
+	// authoritative after a restart or when another node handles the request.
+	z.mpCache.Range(func(uploadID string, _ MultipartInfo) bool {
+		z.mpCache.Delete(uploadID)
+		return true
+	})
+
+	all, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, all, objects...)
+	if all.IsTruncated || all.NextKeyMarker != "" || all.NextUploadIDMarker != "" {
+		t.Fatalf("complete listing has truncation state: %+v", all)
+	}
+
+	first, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, first, objects[0])
+	if !first.IsTruncated || first.NextKeyMarker != objects[0] || first.NextUploadIDMarker != uploadIDs[objects[0]] {
+		t.Fatalf("first page markers = (%q, %q, %t), want (%q, %q, true)",
+			first.NextKeyMarker, first.NextUploadIDMarker, first.IsTruncated,
+			objects[0], uploadIDs[objects[0]])
+	}
+
+	rest, err := z.ListMultipartUploads(t.Context(), bucket, "", first.NextKeyMarker, first.NextUploadIDMarker, "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, rest, objects[1:]...)
+
+	afterKey, err := z.ListMultipartUploads(t.Context(), bucket, "", objects[1], "", "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, afterKey, objects[2:]...)
+
+	prefixed, err := z.ListMultipartUploads(t.Context(), bucket, "t/", "", "", "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, prefixed, objects[:3]...)
+
+	nested, err := z.ListMultipartUploads(t.Context(), bucket, "t/a_b/", "", "", "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, nested, objects[:2]...)
+
+	grouped, err := z.ListMultipartUploads(t.Context(), bucket, "t/", "", "", SlashSeparator, 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, grouped)
+	if want := []string{"t/a_b/", "t/c_d/"}; !slices.Equal(grouped.CommonPrefixes, want) {
+		t.Fatalf("common prefixes = %v, want %v", grouped.CommonPrefixes, want)
+	}
+
+	groupPage, err := z.ListMultipartUploads(t.Context(), bucket, "t/", "", "", SlashSeparator, 1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if want := []string{"t/a_b/"}; !slices.Equal(groupPage.CommonPrefixes, want) {
+		t.Fatalf("first common-prefix page = %v, want %v", groupPage.CommonPrefixes, want)
+	}
+	if !groupPage.IsTruncated || groupPage.NextKeyMarker != "t/a_b/" || groupPage.NextUploadIDMarker != "" {
+		t.Fatalf("common-prefix page markers = (%q, %q, %t)",
+			groupPage.NextKeyMarker, groupPage.NextUploadIDMarker, groupPage.IsTruncated)
+	}
+
+	groupRest, err := z.ListMultipartUploads(t.Context(), bucket, "t/", groupPage.NextKeyMarker, "", SlashSeparator, 1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if want := []string{"t/c_d/"}; !slices.Equal(groupRest.CommonPrefixes, want) {
+		t.Fatalf("second common-prefix page = %v, want %v", groupRest.CommonPrefixes, want)
+	}
+	if groupRest.IsTruncated {
+		t.Fatalf("last common-prefix page is truncated: %+v", groupRest)
+	}
+
+	part, err := z.PutObjectPart(t.Context(), bucket, objects[0], uploadIDs[objects[0]], 1,
+		mustGetPutObjReader(t, bytes.NewBufferString("part"), 4, "", ""), ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	completed, err := z.CompleteMultipartUpload(t.Context(), bucket, objects[0], uploadIDs[objects[0]],
+		[]CompletePart{{PartNumber: 1, ETag: part.ETag}}, ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, key := range []string{multipartMetaBucket, multipartMetaObject} {
+		if _, ok := completed.UserDefined[key]; ok {
+			t.Errorf("completed object retained upload-only metadata %q", key)
+		}
+	}
+
+	// Simulate an upload written by a pre-upgrade server. Its key cannot be
+	// recovered by scanning the hashed namespace. Strict listing must fail;
+	// the old path is available only through an explicit migration setting.
+	legacyObject := objects[1]
+	er := sets.getHashedSet(legacyObject)
+	fi, metadata, err := er.checkUploadIDExists(t.Context(), bucket, legacyObject, uploadIDs[legacyObject], true)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for i := range metadata {
+		delete(metadata[i].Metadata, multipartMetaBucket)
+		delete(metadata[i].Metadata, multipartMetaObject)
+	}
+	if _, err = writeAllMetadata(t.Context(), er.getDisks(), bucket, minioMetaMultipartBucket,
+		er.getUploadIDDir(bucket, legacyObject, uploadIDs[legacyObject]), metadata, fi.WriteQuorum(er.defaultWQuorum())); err != nil {
+		t.Fatal(err)
+	}
+	_, err = z.ListMultipartUploads(t.Context(), bucket, legacyObject, "", "", "", 100)
+	if !errors.Is(err, errMultipartListingLegacy) {
+		t.Fatalf("legacy strict listing: %v", err)
+	}
+	setMultipartListingTestMode(t, true)
+	legacy, err := z.ListMultipartUploads(t.Context(), bucket, legacyObject, "", "", "", 100)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, legacy, legacyObject)
+}
+
+func TestPaginateMultipartUploads(t *testing.T) {
+	base := time.Unix(100, 0)
+	id1, id2, id3 := multipartListingTestID(base, 1), multipartListingTestID(base.Add(time.Second), 2), multipartListingTestID(base, 3)
+	uploads := []MultipartInfo{
+		{Bucket: "bucket", Object: "b", UploadID: id3, Initiated: base},
+		{Bucket: "bucket", Object: "a", UploadID: id2, Initiated: base.Add(time.Second)},
+		{Bucket: "bucket", Object: "a", UploadID: id1, Initiated: base},
+		{Bucket: "bucket", Object: "a", UploadID: id1, Initiated: base}, // duplicate discovery
+	}
+
+	first := paginateMultipartUploads(uploads, "", "", "", "", 1)
+	requireMultipartUploadKeys(t, first, "a")
+	if !first.IsTruncated || first.NextKeyMarker != "a" || first.NextUploadIDMarker != id1 {
+		t.Fatalf("first page = %+v", first)
+	}
+
+	second := paginateMultipartUploads(uploads, "", first.NextKeyMarker, first.NextUploadIDMarker, "", 1)
+	if len(second.Uploads) != 1 || second.Uploads[0].Object != "a" || second.Uploads[0].UploadID != id2 {
+		t.Fatalf("second page uploads = %+v", second.Uploads)
+	}
+	if !second.IsTruncated || second.NextKeyMarker != "a" || second.NextUploadIDMarker != id2 {
+		t.Fatalf("second page = %+v", second)
+	}
+
+	last := paginateMultipartUploads(uploads, "", second.NextKeyMarker, second.NextUploadIDMarker, "", 1)
+	requireMultipartUploadKeys(t, last, "b")
+	if last.IsTruncated || last.NextKeyMarker != "" || last.NextUploadIDMarker != "" {
+		t.Fatalf("last page = %+v", last)
+	}
+
+	missingUploadMarker := paginateMultipartUploads(uploads, "", "a", multipartListingTestID(base.Add(2*time.Second), 4), "", 10)
+	requireMultipartUploadKeys(t, missingUploadMarker, "b")
+
+	if err := checkListMultipartArgs(t.Context(), "bucket", "", "", "not-base64=", ""); err != nil {
+		t.Fatalf("upload-id-marker without key-marker must be ignored: %v", err)
+	}
+
+	overLimit := make([]MultipartInfo, maxUploadsList+1)
+	for i := range overLimit {
+		overLimit[i] = MultipartInfo{Bucket: "bucket", Object: fmt.Sprintf("%04d", i), UploadID: fmt.Sprint(i)}
+	}
+	capped := paginateMultipartUploads(overLimit, "", "", "", "", maxUploadsList+1)
+	if capped.MaxUploads != maxUploadsList || len(capped.Uploads) != maxUploadsList || !capped.IsTruncated {
+		t.Fatalf("over-limit page = MaxUploads %d, uploads %d, truncated %t",
+			capped.MaxUploads, len(capped.Uploads), capped.IsTruncated)
+	}
+}
+
+func TestListMultipartUploadsGlobalPageAcrossPools(t *testing.T) {
+	z, bucket := consistencyPools(t)
+	setMultipartListingTestMode(t, false)
+	objects := []string{"a/one", "b/two", "c/three", "d/four"}
+	for i, object := range objects {
+		if _, err := z.serverPools[i%len(z.serverPools)].NewMultipartUpload(t.Context(), bucket, object, ObjectOptions{}); err != nil {
+			t.Fatalf("NewMultipartUpload(%q): %v", object, err)
+		}
+	}
+	z.mpCache.Range(func(uploadID string, _ MultipartInfo) bool {
+		z.mpCache.Delete(uploadID)
+		return true
+	})
+
+	page, err := z.ListMultipartUploads(t.Context(), bucket, "", "", "", "", 2)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, page, objects[:2]...)
+	if !page.IsTruncated || page.NextKeyMarker != objects[1] {
+		t.Fatalf("first global page = %+v", page)
+	}
+
+	rest, err := z.ListMultipartUploads(t.Context(), bucket, "", page.NextKeyMarker, page.NextUploadIDMarker, "", 2)
+	if err != nil {
+		t.Fatal(err)
+	}
+	requireMultipartUploadKeys(t, rest, objects[2:]...)
+	if rest.IsTruncated {
+		t.Fatalf("last global page is truncated: %+v", rest)
+	}
+}
@@ -0,0 +1,433 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"context"
+	"encoding/xml"
+	"fmt"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"slices"
+	"sync"
+	"testing"
+	"time"
+
+	madmin "github.com/minio/madmin-go/v3"
+)
+
+// Hold nine completed disk walks while a PUT overwrites one name. The other
+// seven walks start after that PUT completes. Both generations come from real
+// disk walkers and valid PUT metadata; only the reader schedule is controlled.
+type nullQuorumWalkSchedule struct {
+	bucket    string
+	oldReady  chan struct{}
+	resume    chan struct{}
+	mu        sync.Mutex
+	snapshots [16][]byte
+}
+
+type nullQuorumWalkDisk struct {
+	StorageAPI
+	schedule *nullQuorumWalkSchedule
+	index    int
+}
+
+func (d *nullQuorumWalkDisk) WalkDir(ctx context.Context, opts WalkDirOptions, out io.Writer) error {
+	if opts.Bucket != d.schedule.bucket {
+		return d.StorageAPI.WalkDir(ctx, opts, out)
+	}
+	var stream bytes.Buffer
+	if d.index < 9 {
+		if err := d.StorageAPI.WalkDir(ctx, opts, &stream); err != nil {
+			return err
+		}
+		select {
+		case d.schedule.oldReady <- struct{}{}:
+		case <-ctx.Done():
+			return ctx.Err()
+		}
+	}
+	select {
+	case <-d.schedule.resume:
+	case <-ctx.Done():
+		return ctx.Err()
+	}
+	if d.index >= 9 {
+		if err := d.StorageAPI.WalkDir(ctx, opts, &stream); err != nil {
+			return err
+		}
+	}
+	d.schedule.mu.Lock()
+	d.schedule.snapshots[d.index] = bytes.Clone(stream.Bytes())
+	d.schedule.mu.Unlock()
+	_, err := out.Write(stream.Bytes())
+	return err
+}
+
+func nullQuorumBackend(t *testing.T) (*erasureServerPools, string, http.Handler) {
+	t.Helper()
+	ctx, cancel := context.WithCancel(t.Context())
+	obj, dirs, err := prepareErasure16(ctx)
+	if err != nil {
+		cancel()
+		t.Fatal(err)
+	}
+	z := obj.(*erasureServerPools)
+	previous := newObjectLayerFn()
+	setObjectLayer(z)
+	t.Cleanup(func() {
+		cancel()
+		z.Shutdown(context.Background())
+		removeRoots(dirs)
+		setObjectLayer(previous)
+	})
+	bucket, router, err := initAPIHandlerTest(ctx, z, nil, MakeBucketOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	return z, bucket, router
+}
+
+func nullQuorumRequest(t *testing.T, router http.Handler, method, target string) *httptest.ResponseRecorder {
+	t.Helper()
+	req, err := newTestSignedRequestV4(method, target, 0, nil, globalActiveCred.AccessKey, globalActiveCred.SecretKey, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	rec := httptest.NewRecorder()
+	router.ServeHTTP(rec, req.WithContext(t.Context()))
+	return rec
+}
+
+func TestListObjectsSingleNullQuorumHTTP(t *testing.T) {
+	z, bucket, router := nullQuorumBackend(t)
+	oldBody := bytes.Repeat([]byte{'a'}, 8192)
+	newBody := bytes.Repeat([]byte{'b'}, 8192)
+	oldTime := time.Now().UTC().Add(-time.Hour)
+	newTime := oldTime.Add(time.Minute)
+	var want []string
+	for i := range 16 {
+		name := fmt.Sprintf("fixed/%02d", i)
+		want = append(want, name)
+		_, err := z.PutObject(t.Context(), bucket, name, mustGetPutObjReader(t, bytes.NewReader(oldBody), int64(len(oldBody)), "", ""), ObjectOptions{MTime: oldTime})
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+	const overwritten = "fixed/10"
+	schedule := &nullQuorumWalkSchedule{bucket: bucket, oldReady: make(chan struct{}, 16), resume: make(chan struct{})}
+	set := z.serverPools[0].sets[0]
+	disks := set.getDisks()
+	wrapped := make([]StorageAPI, len(disks))
+	for i, disk := range disks {
+		wrapped[i] = &nullQuorumWalkDisk{StorageAPI: disk, schedule: schedule, index: i}
+	}
+	set.getDisks = func() []StorageAPI { return wrapped }
+	if globalAPIConfig.getListQuorum() != "strict" {
+		t.Fatal("test requires strict listing across all sixteen disks")
+	}
+	writeDone := make(chan error, 1)
+	putReader := mustGetPutObjReader(t, bytes.NewReader(newBody), int64(len(newBody)), "", "")
+	go func() {
+		defer close(schedule.resume)
+		ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+		defer cancel()
+		for range 9 {
+			select {
+			case <-schedule.oldReady:
+			case <-ctx.Done():
+				writeDone <- ctx.Err()
+				return
+			}
+		}
+		_, err := z.PutObject(ctx, bucket, overwritten, putReader, ObjectOptions{MTime: newTime})
+		writeDone <- err
+	}()
+	rec := nullQuorumRequest(t, router, http.MethodGet, getListObjectsV2URL("", bucket, "fixed/", "1000", "", "", ""))
+	if err := <-writeDone; err != nil {
+		t.Fatal(err)
+	}
+	var list ListObjectsV2Response
+	if rec.Code != http.StatusOK || xml.Unmarshal(rec.Body.Bytes(), &list) != nil {
+		t.Fatalf("LIST: %d %s", rec.Code, rec.Body.String())
+	}
+	var got []string
+	for _, object := range list.Contents {
+		got = append(got, object.Key)
+	}
+	t.Logf("LIST HTTP=%d KeyCount=%d IsTruncated=%v keys=%v", rec.Code, list.KeyCount, list.IsTruncated, got)
+	if list.KeyCount != 16 || list.IsTruncated || !slices.Equal(got, want) {
+		t.Errorf("LIST omitted a name during overwrite: got %v, want %v", got, want)
+	}
+
+	// Verify the actual reader inputs, including shape and EC, instead of
+	// assuming the scheduling barrier produced the intended resolver case.
+	schedule.mu.Lock()
+	snapshots := schedule.snapshots
+	schedule.mu.Unlock()
+	for i, snapshot := range snapshots {
+		reader := newMetacacheReader(bytes.NewReader(snapshot))
+		var found bool
+		for {
+			entry, err := reader.next()
+			if err == io.EOF {
+				break
+			}
+			if err != nil {
+				t.Fatal(err)
+			}
+			if entry.name != overwritten {
+				continue
+			}
+			xl, err := entry.xlmeta()
+			if err != nil || len(xl.versions) != 1 {
+				t.Fatalf("disk %d: invalid input shape: %v", i, err)
+			}
+			header := xl.versions[0].header
+			wantTime := oldTime
+			if i >= 9 {
+				wantTime = newTime
+			}
+			if header.ModTime != wantTime.UnixNano() || header.VersionID != [16]byte{} || header.Type != ObjectType || header.FreeVersion() {
+				t.Fatalf("disk %d: unexpected header %v", i, header)
+			}
+			t.Logf("disk=%d versions=1 header=%v", i, header)
+			found = true
+		}
+		reader.Close()
+		if !found {
+			t.Fatalf("disk %d did not emit the overwritten name", i)
+		}
+	}
+	get := nullQuorumRequest(t, router, http.MethodGet, getGetObjectURL("", bucket, overwritten))
+	head := nullQuorumRequest(t, router, http.MethodHead, getGetObjectURL("", bucket, overwritten))
+	t.Logf("completed PUT readback: GET=%d bytes=%d HEAD=%d length=%s", get.Code, get.Body.Len(), head.Code, head.Header().Get("Content-Length"))
+	if get.Code != http.StatusOK || !bytes.Equal(get.Body.Bytes(), newBody) || head.Code != http.StatusOK || head.Header().Get("Content-Length") != "8192" {
+		t.Fatalf("completed PUT was not readable: GET=%d HEAD=%d", get.Code, head.Code)
+	}
+}
+
+func TestSingleNullQuorumReadAndConsumers(t *testing.T) {
+	z, bucket, router := nullQuorumBackend(t)
+	set := z.serverPools[0].sets[0]
+	disks := set.getDisks()
+	const object = "object"
+	oldBody, newBody := bytes.Repeat([]byte{'a'}, 8192), bytes.Repeat([]byte{'b'}, 8192)
+	oldTime := time.Now().UTC().Add(-time.Hour)
+	put := func(body []byte, modTime time.Time) {
+		t.Helper()
+		_, err := z.PutObject(t.Context(), bucket, object, mustGetPutObjReader(t, bytes.NewReader(body), int64(len(body)), "", ""), ObjectOptions{MTime: modTime})
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+	put(oldBody, oldTime)
+	oldMeta := make([][]byte, len(disks))
+	for i, disk := range disks {
+		var err error
+		oldMeta[i], err = disk.ReadAll(t.Context(), bucket, object+"/"+xlStorageFormatFile)
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+	put(newBody, oldTime.Add(time.Minute))
+	newMinorityMeta := mustReadNullQuorumMeta(t, disks[14], bucket, object)
+	// Model an incomplete overwrite using real inline shards: fifteen old
+	// copies retain a read quorum; the final disk contains the newer minority.
+	for i := range 15 {
+		if err := disks[i].WriteAll(t.Context(), bucket, object+"/"+xlStorageFormatFile, oldMeta[i]); err != nil {
+			t.Fatal(err)
+		}
+	}
+	checkRead := func(body []byte, modTime time.Time) {
+		t.Helper()
+		get := nullQuorumRequest(t, router, http.MethodGet, getGetObjectURL("", bucket, object))
+		head := nullQuorumRequest(t, router, http.MethodHead, getGetObjectURL("", bucket, object))
+		if get.Code != 200 || !bytes.Equal(get.Body.Bytes(), body) || head.Code != 200 || head.Header().Get("Content-Length") != "8192" {
+			t.Fatalf("GET/HEAD lost readable quorum: GET=%d HEAD=%d body=%q", get.Code, head.Code, get.Body.String())
+		}
+		oi, err := z.GetObjectInfo(t.Context(), bucket, object, ObjectOptions{})
+		if err != nil || !oi.ModTime.Equal(modTime) {
+			t.Fatalf("wrong readable generation: %v, %v", oi.ModTime, err)
+		}
+	}
+	checkRead(oldBody, oldTime)
+	// Reading must not reinterpret the minority as absent or delete it.
+	minority, err := disks[15].ReadVersion(t.Context(), "", bucket, object, "", ReadOptions{})
+	if err != nil || !minority.ModTime.Equal(oldTime.Add(time.Minute)) {
+		t.Fatalf("read mutated the minority: %+v, %v", minority, err)
+	}
+	for _, kind := range []string{"rebalance", "decommission"} {
+		t.Run(kind, func(t *testing.T) {
+			var names []string
+			consume := func(entry metaCacheEntry) {
+				versions, err := entry.fileInfoVersions(bucket)
+				if err != nil || len(versions.Versions) != 1 || !versions.Versions[0].ModTime.Equal(oldTime) {
+					t.Errorf("invalid migration metadata: %+v, %v", versions, err)
+					return
+				}
+				// Migration reopens the selected name/version on all drives;
+				// it must not treat the selected listing metadata as disk agreement.
+				reader, err := set.GetObjectNInfo(t.Context(), bucket, entry.name, nil, nil, ObjectOptions{VersionID: nullVersionID, NoLock: true, NoDecryption: true})
+				if err != nil {
+					t.Error(err)
+					return
+				}
+				body, err := io.ReadAll(reader)
+				reader.Close()
+				if err != nil || !bytes.Equal(body, oldBody) {
+					t.Errorf("migration could not read selected object data: %v", err)
+				}
+				names = append(names, entry.name)
+			}
+			var err error
+			if kind == "rebalance" {
+				err = set.listObjectsToRebalance(t.Context(), bucket, consume)
+			} else {
+				err = set.listObjectsToDecommission(t.Context(), decomBucketInfo{Name: bucket}, consume)
+			}
+			if err != nil || !slices.Equal(names, []string{object}) {
+				t.Fatalf("migration listing dropped the quorum: %v, %v", names, err)
+			}
+		})
+	}
+	for _, latestOnly := range []bool{false, true} {
+		results := make(chan itemOrErr[ObjectInfo], 16)
+		if err := z.Walk(t.Context(), bucket, "", results, WalkOptions{AskDisks: "strict", LatestOnly: latestOnly}); err != nil {
+			t.Fatal(err)
+		}
+		var names []string
+		for result := range results {
+			if result.Err != nil || !result.Item.ModTime.Equal(oldTime) {
+				t.Fatalf("Walk returned invalid metadata: %+v", result)
+			}
+			names = append(names, result.Item.Name)
+		}
+		if !slices.Equal(names, []string{object}) {
+			t.Fatalf("Walk dropped the quorum: %v", names)
+		}
+	}
+
+	// Exercise the scanner's actual abandoned-child path. Make the scan drive
+	// miss the object, while keeping an older quorum and a newer minority on
+	// the remaining disks. Its queued heal must read all disks and repair both.
+	if err := disks[14].WriteAll(t.Context(), bucket, object+"/"+xlStorageFormatFile, newMinorityMeta); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.RemoveAll(filepath.Join(disks[15].Endpoint().Path, bucket, object)); err != nil {
+		t.Fatal(err)
+	}
+	scanNullQuorumAbandoned(t, z, bucket, object, oldTime)
+	for i, disk := range disks {
+		fi, err := disk.ReadVersion(t.Context(), "", bucket, object, "", ReadOptions{Healing: true})
+		if err != nil || !fi.ModTime.Equal(oldTime) {
+			t.Fatalf("scanner heal did not reconcile disk %d: %v, %v", i, fi.ModTime, err)
+		}
+	}
+	checkRead(oldBody, oldTime)
+	put(newBody, oldTime.Add(2*time.Minute))
+	checkRead(newBody, oldTime.Add(2*time.Minute))
+}
+
+func mustReadNullQuorumMeta(t *testing.T, disk StorageAPI, bucket, object string) []byte {
+	t.Helper()
+	data, err := disk.ReadAll(t.Context(), bucket, object+"/"+xlStorageFormatFile)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return data
+}
+
+func scanNullQuorumAbandoned(t *testing.T, z *erasureServerPools, bucket, object string, oldTime time.Time) {
+	t.Helper()
+	ctx, cancel := context.WithCancel(t.Context())
+	defer cancel()
+	disks := z.serverPools[0].sets[0].getDisks()
+	seq := newBgHealSequence()
+	defer seq.cancelCtx()
+	previousState, previousRoutine := globalBackgroundHealState, globalBackgroundHealRoutine
+	globalBackgroundHealState = &allHealState{healSeqMap: map[string]*healSequence{"test": seq}}
+	routine := &healRoutine{tasks: make(chan healTask)}
+	globalBackgroundHealRoutine = routine
+	defer func() { globalBackgroundHealState, globalBackgroundHealRoutine = previousState, previousRoutine }()
+	workerDone := make(chan struct{})
+	var healed []string
+	go func() {
+		defer close(workerDone)
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case task := <-routine.tasks:
+				var result madmin.HealResultItem
+				var err error
+				if task.object == "" {
+					result, err = z.HealBucket(ctx, task.bucket, task.opts)
+				} else {
+					healed = append(healed, task.object+"/"+task.versionID)
+					result, err = z.HealObject(ctx, task.bucket, task.object, task.versionID, task.opts)
+				}
+				task.respCh <- healResult{result: result, err: err}
+			}
+		}
+	}()
+	cache := dataUsageCache{Info: dataUsageCacheInfo{Name: bucket}}
+	cache.replace(bucket, "", dataUsageEntry{})
+	cache.replace(bucket+"/"+object, bucket, dataUsageEntry{Size: 8192, Objects: 1, Versions: 1})
+	scanner := folderScanner{
+		root: disks[15].Endpoint().Path, oldCache: cache,
+		newCache: dataUsageCache{Info: cache.Info}, updateCache: dataUsageCache{Info: cache.Info},
+		disks: disks, disksQuorum: 8, healObjectSelect: 1,
+		weSleep: func() bool { return false }, shouldHeal: func() bool { return true }, updateCurrentPath: func(string) {},
+		getSize: func(item scannerItem) (sizeSummary, error) {
+			if filepath.Base(item.Path) != xlStorageFormatFile {
+				return sizeSummary{}, errSkipFile
+			}
+			data, err := os.ReadFile(item.Path)
+			if err != nil {
+				return sizeSummary{}, err
+			}
+			var xl xlMetaV2
+			if err := xl.Load(data); err != nil {
+				return sizeSummary{}, err
+			}
+			fi, err := xl.ToFileInfo(bucket, object, "", false, false)
+			if err != nil || !fi.ModTime.Equal(oldTime) {
+				return sizeSummary{}, fmt.Errorf("scanner re-read wrong generation: %v, %v", fi.ModTime, err)
+			}
+			return sizeSummary{totalSize: fi.Size, versions: 1}, nil
+		},
+	}
+	var usage dataUsageEntry
+	err := scanner.scanFolder(ctx, cachedFolder{name: bucket, objectHealProbDiv: 1}, &usage)
+	cancel()
+	<-workerDone
+	if err != nil || len(healed) != 1 || healed[0] != object+"/"+nullVersionID && healed[0] != object+"/" {
+		t.Fatalf("scanner did not queue a name-based heal: %v, %v", healed, err)
+	}
+	flat := scanner.newCache.sizeRecursive(bucket)
+	if flat == nil || flat.Objects != 1 || flat.Size != 8192 {
+		t.Fatalf("scanner lost the healed object from usage: %+v", flat)
+	}
+	t.Logf("scanner queued %v; healed old quorum, repaired minority/missing copies, and retained one 8192-byte object", healed)
+}
@@ -69,6 +69,8 @@ func loadCPUMetrics(ctx context.Context, m MetricValues, c *metricsCache) error

 	// metrics-resource.go runs a job to collect resource metrics including their Avg values and
 	// stores them in resourceMetricsMap. We can use it to get the Avg values of CPU idle and IOWait.
+	resourceMetricsMapMu.RLock()
+	defer resourceMetricsMapMu.RUnlock()
 	cpuResourceMetrics, found := resourceMetricsMap[cpuSubsystem]
 	if found {
 		if cpuIdleMetric, ok := cpuResourceMetrics[getResourceKey(cpuIdle, nil)]; ok {
@@ -0,0 +1,185 @@
+// Copyright (c) 2026 Ruohang Feng
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"fmt"
+	"runtime"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio/internal/cachevalue"
+	"github.com/prometheus/client_golang/prometheus"
+	dto "github.com/prometheus/client_model/go"
+	cpustats "github.com/shirou/gopsutil/v3/cpu"
+	"github.com/shirou/gopsutil/v3/load"
+)
+
+// These tests replace global resource metrics and must not run in parallel.
+// Concurrent writers must finish before the fixture is restored.
+func setCPUResourceMetricsForTest(t *testing.T, value map[MetricSubsystem]ResourceMetrics) {
+	t.Helper()
+	resourceMetricsMapMu.Lock()
+	saved := resourceMetricsMap
+	resourceMetricsMap = value
+	resourceMetricsMapMu.Unlock()
+	t.Cleanup(func() {
+		resourceMetricsMapMu.Lock()
+		resourceMetricsMap = saved
+		resourceMetricsMapMu.Unlock()
+	})
+}
+
+func newCPUMetricsTestRegistry() *prometheus.Registry {
+	c := &metricsCache{cpuMetrics: cachevalue.NewFromFunc(time.Hour, cachevalue.Opts{},
+		func(context.Context) (madmin.CPUMetrics, error) {
+			return madmin.CPUMetrics{
+				CPUCount: 4,
+				LoadStat: &load.AvgStat{Load1: 2},
+				TimesStat: &cpustats.TimesStat{
+					User: 10, System: 20, Idle: 60, Iowait: 5, Nice: 3, Steal: 2,
+				},
+			}, nil
+		})}
+	_, _ = c.cpuMetrics.Get() // Warm the cache: the cache does not protect the resource map.
+	g := NewMetricsGroup(systemCPUCollectorPath, []MetricDescriptor{
+		sysCPUAvgIdleMD, sysCPUAvgIOWaitMD, sysCPULoadMD, sysCPULoadPercMD,
+		sysCPUNiceMD, sysCPUStealMD, sysCPUSystemMD, sysCPUUserMD,
+	}, loadCPUMetrics)
+	g.SetCache(c)
+	r := prometheus.NewPedanticRegistry()
+	r.MustRegister(g)
+	return r
+}
+
+func TestLoadCPUMetricsValues(t *testing.T) {
+	for _, tc := range []struct {
+		name string
+		data map[MetricSubsystem]ResourceMetrics
+		want map[string]float64
+	}{
+		{name: "nil-map"},
+		{name: "empty-map", data: map[MetricSubsystem]ResourceMetrics{}},
+		{name: "no-cpu", data: map[MetricSubsystem]ResourceMetrics{memSubsystem: {}}},
+		{name: "nil-cpu", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: nil}},
+		{name: "empty-cpu", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: {}}},
+		{name: "idle-only", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: {
+			getResourceKey(cpuIdle, nil): {Avg: 87.654},
+		}}, want: map[string]float64{"minio_system_cpu_avg_idle": 87.65}},
+		{name: "iowait-only", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: {
+			getResourceKey(cpuIOWait, nil): {Avg: 1.236},
+		}}, want: map[string]float64{"minio_system_cpu_avg_iowait": 1.24}},
+		{name: "both-rounded", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: {
+			getResourceKey(cpuIdle, nil):   {Avg: 87.654},
+			getResourceKey(cpuIOWait, nil): {Avg: 1.236},
+		}}, want: map[string]float64{"minio_system_cpu_avg_idle": 87.65, "minio_system_cpu_avg_iowait": 1.24}},
+		{name: "zero-keeps-existing-omission", data: map[MetricSubsystem]ResourceMetrics{cpuSubsystem: {
+			getResourceKey(cpuIdle, nil):   {Avg: 0},
+			getResourceKey(cpuIOWait, nil): {Avg: 0},
+		}}},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			setCPUResourceMetricsForTest(t, tc.data)
+			families, err := newCPUMetricsTestRegistry().Gather()
+			if err != nil {
+				t.Fatal(err)
+			}
+			want := map[string]float64{
+				"minio_system_cpu_load": 2, "minio_system_cpu_load_perc": 50,
+				"minio_system_cpu_nice": 3, "minio_system_cpu_steal": 2,
+				"minio_system_cpu_system": 20, "minio_system_cpu_user": 10,
+			}
+			for k, v := range tc.want {
+				want[k] = v
+			}
+			if len(families) != len(want) {
+				t.Fatalf("got %d families, want %d", len(families), len(want))
+			}
+			for _, family := range families {
+				v, ok := want[family.GetName()]
+				if !ok || family.GetType() != dto.MetricType_GAUGE || len(family.Metric) != 1 || family.Metric[0].GetGauge().GetValue() != v {
+					t.Errorf("unexpected family: %v (want %v)", family, want)
+				}
+			}
+		})
+	}
+}
+
+func TestLoadCPUMetricsConcurrentUpdate(t *testing.T) {
+	for _, subsystem := range []MetricSubsystem{cpuSubsystem, memSubsystem} {
+		for _, readers := range []int{1, 4} {
+			t.Run(fmt.Sprintf("writer-%s/readers-%d", subsystem, readers), func(t *testing.T) {
+				setCPUResourceMetricsForTest(t, map[MetricSubsystem]ResourceMetrics{})
+				updateResourceMetrics(cpuSubsystem, cpuIdle, 80, nil, false)
+				updateResourceMetrics(cpuSubsystem, cpuIOWait, 5, nil, false)
+				r := newCPUMetricsTestRegistry()
+				stop := make(chan struct{})
+				done := make(chan struct{})
+				ready := make(chan struct{})
+				var updates atomic.Uint64
+				go func() {
+					defer close(done)
+					if subsystem == cpuSubsystem {
+						updateResourceMetrics(subsystem, cpuIdle, 80, nil, false)
+					} else {
+						updateResourceMetrics(subsystem, memUsed, 1024, nil, false)
+					}
+					close(ready)
+					for {
+						select {
+						case <-stop:
+							return
+						default:
+						}
+						if subsystem == cpuSubsystem {
+							updateResourceMetrics(subsystem, cpuIdle, 80, nil, false)
+							updateResourceMetrics(subsystem, cpuIOWait, 5, nil, false)
+						} else {
+							updateResourceMetrics(subsystem, memUsed, 1024, nil, false)
+						}
+						updates.Add(1)
+						runtime.Gosched()
+					}
+				}()
+				<-ready
+				defer func() { close(stop); <-done }()
+				var wg sync.WaitGroup
+				for range readers {
+					wg.Add(1)
+					go func() {
+						defer wg.Done()
+						for range 1000 {
+							families, err := r.Gather()
+							if err != nil || len(families) != 8 {
+								t.Errorf("Gather: families=%d, error=%v", len(families), err)
+								return
+							}
+						}
+					}()
+				}
+				wg.Wait()
+				if updates.Load() == 0 {
+					t.Fatal("writer made no progress")
+				}
+				t.Logf("completed %d gathers with %d concurrent update iterations", readers*1000, updates.Load())
+			})
+		}
+	}
+}
@@ -20,6 +20,7 @@ package cmd
 import (
 	"context"
 	"encoding/base64"
+	"errors"
 	"runtime"
 	"strings"

@@ -83,7 +84,8 @@ func checkListMultipartArgs(ctx context.Context, bucket, prefix, keyMarker, uplo
 	if err := checkListObjsArgs(ctx, bucket, prefix, keyMarker); err != nil {
 		return err
 	}
-	if uploadIDMarker != "" {
+	// S3 ignores upload-id-marker when key-marker is absent.
+	if uploadIDMarker != "" && keyMarker != "" {
 		if HasSuffix(keyMarker, SlashSeparator) {
 			return InvalidUploadIDKeyCombination{
 				UploadIDMarker: uploadIDMarker,
@@ -96,6 +98,9 @@ func checkListMultipartArgs(ctx context.Context, bucket, prefix, keyMarker, uplo
 				UploadID: uploadIDMarker,
 			}
 		}
+		if _, ok := multipartMarkerTime(uploadIDMarker); !ok {
+			return InvalidArgument{Bucket: bucket, Object: keyMarker, Err: errors.New("upload-id-marker must contain a native multipart upload ID")}
+		}
 	}
 	return nil
 }
@@ -82,17 +82,26 @@ type markerRecoveryTarget struct {
 func replicationTestCapacity(obj ObjectLayer) func() {
 	var restore []func()
 	for _, pool := range obj.(*erasureServerPools).serverPools {
+		pool.erasureDisksMu.Lock()
 		for _, set := range pool.sets {
-			original := set.getDisks
-			disks := append([]StorageAPI(nil), original()...)
+			original := pool.erasureDisks[set.setIndex]
+			disks := append([]StorageAPI(nil), original...)
 			for i, disk := range disks {
 				if disk != nil {
 					disks[i] = tagTestCapacityDisk{StorageAPI: disk}
 				}
 			}
-			set.getDisks = func() []StorageAPI { return disks }
-			restore = append(restore, func() { set.getDisks = original })
+			// GetDisks copies this list under the same mutex. Keep its function
+			// stable while background IAM scans use the fixture, both when
+			// installing the adapter and when restoring the original disks.
+			pool.erasureDisks[set.setIndex] = disks
+			restore = append(restore, func() {
+				pool.erasureDisksMu.Lock()
+				pool.erasureDisks[set.setIndex] = original
+				pool.erasureDisksMu.Unlock()
+			})
 		}
+		pool.erasureDisksMu.Unlock()
 	}
 	return func() {
 		for _, fn := range restore {
@@ -101,6 +110,34 @@ func replicationTestCapacity(obj ObjectLayer) func() {
 	}
 }

+func TestReplicationCapacityConcurrentIAM(t *testing.T) {
+	z, _ := consistencyPools(t)
+	if _, _, err := initAPIHandlerTest(t.Context(), z, nil, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+	defer cancel()
+	iam := globalIAMSys
+	started, finished := make(chan struct{}), make(chan error, 1)
+	go func() {
+		close(started)
+		for range 20 {
+			if err := iam.Load(ctx, false); err != nil {
+				finished <- err
+				return
+			}
+		}
+		finished <- nil
+	}()
+	<-started
+	for range 5000 {
+		replicationTestCapacity(z)()
+	}
+	if err := <-finished; err != nil {
+		t.Fatal(err)
+	}
+}
+
 func testReplicationMRFMarkerRecovery(t *testing.T, obj ObjectLayer, backend, bucket string, router http.Handler, creds auth.Credentials, tc markerRecoveryCase) {
 	ctx, cancel := context.WithCancel(t.Context())
 	defer cancel()
@@ -0,0 +1,264 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio-go/v7"
+	"github.com/minio/minio/internal/auth"
+	"github.com/minio/minio/internal/bucket/replication"
+	xhttp "github.com/minio/minio/internal/http"
+	"github.com/minio/minio/internal/once"
+)
+
+type staleCreationCase struct {
+	name                             string
+	purged, pendingPurge, unreadable bool
+}
+
+// A queued delete-marker creation must be checked against the source marker
+// under the replication lock before it is sent. Between queueing (DELETE
+// handler, GET/HEAD/LIST heal, scanner, MRF) and sending, a user purge of the
+// same marker can complete and reach the targets; the stale creation would
+// then recreate the marker there.
+func TestReplicationDeleteMarkerCreationRevalidated(t *testing.T) {
+	for _, tc := range []staleCreationCase{
+		{name: "current"},
+		{name: "purged", purged: true},
+		{name: "purge-pending", pendingPurge: true},
+		{name: "unreadable", unreadable: true},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, endpoints: []string{"DeleteObject"}, objAPITest: func(obj ObjectLayer, backend, bucket string, router http.Handler, creds auth.Credentials, t *testing.T) {
+				testReplicationDeleteMarkerCreationRevalidated(t, obj, backend, bucket, router, creds, tc)
+			}})
+		})
+	}
+}
+
+type staleCreationTarget struct {
+	arn, bucket string
+	creations   atomic.Int32
+	purges      atomic.Int32
+}
+
+func testReplicationDeleteMarkerCreationRevalidated(t *testing.T, obj ObjectLayer, backend, bucket string, router http.Handler, creds auth.Credentials, tc staleCreationCase) {
+	ctx, cancel := context.WithCancel(t.Context())
+	defer cancel()
+	defer replicationTestCapacity(obj)()
+	stats := NewReplicationStats(ctx, nil)
+	oldStats := globalReplicationStats.Swap(stats)
+	defer globalReplicationStats.Store(oldStats)
+	oldPool := globalReplicationPool
+	defer func() { globalReplicationPool = oldPool }()
+	const name = "marker"
+	if _, err := globalBucketMetadataSys.Update(ctx, bucket, bucketVersioningConfig, enabledBucketVersioningConfig); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := obj.PutObject(ctx, bucket, name, mustGetPutObjReader(t, bytes.NewReader([]byte("data")), 4, "", ""), ObjectOptions{Versioned: true}); err != nil {
+		t.Fatal(err)
+	}
+	target := &staleCreationTarget{arn: "arn:minio:replication::" + mustGetUUID() + ":bucket", bucket: getRandomBucketName()}
+	if err := obj.MakeBucket(ctx, target.bucket, MakeBucketOptions{VersioningEnabled: true}); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := obj.PutObject(ctx, target.bucket, name, mustGetPutObjReader(t, bytes.NewReader([]byte("data")), 4, "", ""), ObjectOptions{Versioned: true}); err != nil {
+		t.Fatal(err)
+	}
+	// The remote behaves like a real receiver: a marker lookup and a
+	// replicated DELETE applied to the target bucket on the same fixture.
+	remote := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		opts := ObjectOptions{VersionID: r.URL.Query().Get("versionId"), Versioned: true}
+		switch r.Method {
+		case http.MethodHead:
+			oi, err := obj.GetObjectInfo(r.Context(), target.bucket, name, opts)
+			if oi.DeleteMarker {
+				w.Header().Set(xhttp.AmzDeleteMarker, "true")
+				w.Header().Set(xhttp.AmzVersionID, oi.VersionID)
+			}
+			if err != nil {
+				writeErrorResponseHeadersOnly(w, toAPIError(r.Context(), err))
+				return
+			}
+			w.WriteHeader(http.StatusOK)
+		case http.MethodDelete:
+			opts.DeleteMarker = r.Header.Get(xhttp.MinIOSourceDeleteMarker) == "true"
+			if opts.DeleteMarker {
+				target.creations.Add(1)
+			} else {
+				target.purges.Add(1)
+			}
+			opts.ReplicationRequest = true
+			opts.SetReplicaStatus(replication.Replica)
+			if _, err := obj.DeleteObject(r.Context(), target.bucket, name, opts); err != nil && !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+				writeErrorResponse(r.Context(), w, toAPIError(r.Context(), err), r.URL)
+				return
+			}
+			w.WriteHeader(http.StatusNoContent)
+		default:
+			t.Errorf("unexpected remote method %s", r.Method)
+		}
+	}))
+	defer remote.Close()
+	client, err := minio.New(strings.TrimPrefix(remote.URL, "http://"), &minio.Options{Region: "us-east-1", MaxRetries: 1})
+	if err != nil {
+		t.Fatal(err)
+	}
+	globalBucketTargetSys.Lock()
+	globalBucketTargetSys.arnRemotesMap[target.arn] = arnTarget{Client: &TargetClient{Client: client, ARN: target.arn, Bucket: target.bucket}, lastRefresh: UTCNow()}
+	globalBucketTargetSys.targetsMap[bucket] = append(globalBucketTargetSys.targetsMap[bucket], madmin.BucketTarget{Arn: target.arn, TargetBucket: target.bucket})
+	globalBucketTargetSys.Unlock()
+	globalBucketTargetSys.hMutex.Lock()
+	globalBucketTargetSys.hc[client.EndpointURL().Host] = epHealth{Online: true}
+	globalBucketTargetSys.hMutex.Unlock()
+	rule := configs[0].Rules[0]
+	rule.Destination = replication.Destination{ARN: target.arn, Bucket: target.bucket}
+	cfg := replication.Config{RoleArn: target.arn, Rules: []replication.Rule{rule}}
+	meta, err := globalBucketMetadataSys.Get(bucket)
+	if err != nil {
+		t.Fatal(err)
+	}
+	meta.replicationConfig = &cfg
+	globalBucketMetadataSys.Set(bucket, meta)
+	p := &ReplicationPool{ctx: ctx, objLayer: obj, workers: []chan ReplicationWorkerOperation{make(chan ReplicationWorkerOperation, 8)}, stats: stats, mrfSaveCh: make(chan MRFReplicateEntry, 8)}
+	globalReplicationPool = once.NewSingleton[ReplicationPool]()
+	globalReplicationPool.Set(p)
+
+	// The real DELETE handler creates the pending marker and queues the
+	// creation task that a worker would later run.
+	req, err := newTestSignedRequestV4(http.MethodDelete, "/"+bucket+"/"+name, 0, nil, creds.AccessKey, creds.SecretKey, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	w := httptest.NewRecorder()
+	router.ServeHTTP(w, req)
+	if w.Code != http.StatusNoContent {
+		t.Fatalf("source DELETE status %d: %s", w.Code, w.Body)
+	}
+	var creation DeletedObjectReplicationInfo
+	select {
+	case op := <-p.workers[0]:
+		creation = op.(DeletedObjectReplicationInfo)
+	case <-time.After(3 * time.Second):
+		t.Fatal("handler queued no creation task")
+	}
+	version := creation.DeleteMarkerVersionID
+	if version == "" || creation.VersionID != "" || !creation.DeleteMarker {
+		t.Fatalf("handler queued a non-creation task: %+v", creation)
+	}
+	before, err := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true})
+	if !isErrMethodNotAllowed(err) || !before.DeleteMarker || before.ReplicationStatus != replication.Pending {
+		t.Fatalf("source marker not pending: %+v %v", before, err)
+	}
+
+	// Meanwhile the user purges the marker through another path. The purge is
+	// either complete (version gone) or still pending on the source.
+	switch {
+	case tc.purged:
+		if _, err := obj.DeleteObject(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true}); err != nil {
+			t.Fatal(err)
+		}
+		if _, err := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true}); !isErrVersionNotFound(err) && !isErrObjectNotFound(err) {
+			t.Fatalf("source purge left the marker: %v", err)
+		}
+	case tc.pendingPurge:
+		opts := ObjectOptions{VersionID: version, Versioned: true, DeleteReplication: ReplicationState{
+			ReplicateDecisionStr:       creation.ReplicationState.ReplicateDecisionStr,
+			VersionPurgeStatusInternal: target.arn + "=PENDING;",
+			PurgeTargets:               map[string]VersionPurgeStatusType{target.arn: replication.VersionPurgePending},
+		}}
+		if _, err := obj.DeleteObject(ctx, bucket, name, opts); err != nil {
+			t.Fatal(err)
+		}
+		oi, err := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true})
+		if !isErrMethodNotAllowed(err) || !oi.DeleteMarker || oi.VersionPurgeStatus != replication.VersionPurgePending {
+			t.Fatalf("source marker not pending purge: %+v %v", oi, err)
+		}
+	}
+	source := obj
+	if tc.unreadable {
+		source = staleLookupLayer{ObjectLayer: obj}
+	}
+	pending, _ := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true})
+
+	result := replicateDelete(ctx, creation, source)
+
+	after, aerr := obj.GetObjectInfo(ctx, bucket, name, ObjectOptions{VersionID: version, Versioned: true})
+	_, terr := obj.GetObjectInfo(ctx, target.bucket, name, ObjectOptions{VersionID: version, Versioned: true})
+	targetHasMarker := isErrMethodNotAllowed(terr)
+	switch {
+	case tc.purged, tc.pendingPurge, tc.unreadable:
+		if len(result.Targets) != 0 || target.creations.Load() != 0 || target.purges.Load() != 0 {
+			t.Fatalf("%s: stale creation was sent: result=%+v creations=%d purges=%d", tc.name, result, target.creations.Load(), target.purges.Load())
+		}
+		if targetHasMarker {
+			t.Fatalf("%s: target marker recreated from the stale creation", tc.name)
+		}
+		if tc.purged {
+			if !isErrVersionNotFound(aerr) && !isErrObjectNotFound(aerr) {
+				t.Fatalf("stale creation resurrected the source marker: %+v %v", after, aerr)
+			}
+		} else if !isErrMethodNotAllowed(aerr) || !after.DeleteMarker ||
+			after.ReplicationStatusInternal != pending.ReplicationStatusInternal ||
+			after.VersionPurgeStatusInternal != pending.VersionPurgeStatusInternal ||
+			after.UserDefined[ReservedMetadataPrefixLower+ReplicationTimestamp] != pending.UserDefined[ReservedMetadataPrefixLower+ReplicationTimestamp] {
+			t.Fatalf("skipped creation rewrote the source marker: before=%+v after=%+v %v", pending, after, aerr)
+		}
+		if tc.unreadable {
+			select {
+			case entry := <-p.mrfSaveCh:
+				if entry.versionID != version || entry.RetryCount != 1 || entry.Bucket != bucket || entry.Object != name {
+					t.Fatalf("unverified creation queued wrong MRF entry: %+v", entry)
+				}
+			default:
+				t.Fatal("unverified source read did not queue a retry")
+			}
+		}
+		if len(p.mrfSaveCh) != 0 {
+			t.Fatalf("%s: unexpected MRF entries queued: %d", tc.name, len(p.mrfSaveCh))
+		}
+	default:
+		if result.ReplicationStatus() != replication.Completed || target.creations.Load() != 1 || !targetHasMarker {
+			t.Fatalf("current creation not replicated: result=%+v creations=%d targetMarker=%v", result, target.creations.Load(), targetHasMarker)
+		}
+		if !isErrMethodNotAllowed(aerr) || !after.DeleteMarker || replicationStatusesMap(after.ReplicationStatusInternal)[target.arn] != replication.Completed {
+			t.Fatalf("source creation state not completed: %+v %v", after, aerr)
+		}
+		if len(p.mrfSaveCh) != 0 {
+			t.Fatal("completed creation queued MRF work")
+		}
+	}
+	t.Logf("%s: %s checked; creations=%d purges=%d", backend, tc.name, target.creations.Load(), target.purges.Load())
+}
+
+// staleLookupLayer cannot confirm the source marker: the read fails without
+// saying whether the version is present.
+type staleLookupLayer struct{ ObjectLayer }
+
+func (staleLookupLayer) GetObjectInfo(context.Context, string, string, ObjectOptions) (ObjectInfo, error) {
+	return ObjectInfo{}, InsufficientReadQuorum{}
+}
@@ -41,15 +41,23 @@ const r5TagStamp = ReservedMetadataPrefixLower + TaggingTimestamp
 func r5Capacity(z *erasureServerPools) func() {
 	var restores []func()
 	for _, pool := range z.serverPools {
+		pool.erasureDisksMu.Lock()
 		for _, set := range pool.sets {
-			old := set.getDisks
-			disks := append([]StorageAPI(nil), old()...)
+			old := pool.erasureDisks[set.setIndex]
+			disks := append([]StorageAPI(nil), old...)
 			for i := range disks {
 				disks[i] = tagTestCapacityDisk{StorageAPI: disks[i]}
 			}
-			set.getDisks = func() []StorageAPI { return disks }
-			restores = append(restores, func() { set.getDisks = old })
+			// GetDisks copies this list under the same mutex. Keep its function
+			// stable while background IAM scans are using the fixture.
+			pool.erasureDisks[set.setIndex] = disks
+			restores = append(restores, func() {
+				pool.erasureDisksMu.Lock()
+				pool.erasureDisks[set.setIndex] = old
+				pool.erasureDisksMu.Unlock()
+			})
 		}
+		pool.erasureDisksMu.Unlock()
 	}
 	return func() {
 		for _, restore := range restores {
@@ -58,6 +66,34 @@ func r5Capacity(z *erasureServerPools) func() {
 	}
 }

+func TestAPITaggingCapacityConcurrentIAM(t *testing.T) {
+	z, _ := consistencyPools(t)
+	if _, _, err := initAPIHandlerTest(t.Context(), z, nil, MakeBucketOptions{}); err != nil {
+		t.Fatal(err)
+	}
+	ctx, cancel := context.WithTimeout(t.Context(), 30*time.Second)
+	defer cancel()
+	iam := globalIAMSys
+	started, finished := make(chan struct{}), make(chan error, 1)
+	go func() {
+		close(started)
+		for range 20 {
+			if err := iam.Load(ctx, false); err != nil {
+				finished <- err
+				return
+			}
+		}
+		finished <- nil
+	}()
+	<-started
+	for range 5000 {
+		r5Capacity(z)()
+	}
+	if err := <-finished; err != nil {
+		t.Fatal(err)
+	}
+}
+
 func r5Request(t *testing.T, router http.Handler, cred auth.Credentials, method, path, body string, headers map[string]string) *httptest.ResponseRecorder {
 	t.Helper()
 	r, err := newTestSignedRequestV4(method, path, int64(len(body)), strings.NewReader(body), cred.AccessKey, cred.SecretKey, headers)
@@ -109,6 +109,15 @@ func testStorageAPIListDir(t *testing.T, storage StorageAPI) {
 			}
 		}
 	}
+	for _, name := range []string{"one", "two", "three"} {
+		if err := storage.AppendFile(t.Context(), "foo", "bounded/"+name, []byte("x")); err != nil {
+			t.Fatal(err)
+		}
+	}
+	entries, err := storage.ListDir(t.Context(), "", "foo", "bounded", 2)
+	if err != nil || len(entries) != 2 {
+		t.Fatalf("ListDir count was lost in storage/RPC path: %v %v", entries, err)
+	}
 }

 func testStorageAPIReadAll(t *testing.T, storage StorageAPI) {
@@ -0,0 +1,101 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"archive/tar"
+	"bytes"
+	"fmt"
+	"io"
+	"os"
+	"sync"
+	"testing"
+	"testing/iotest"
+
+	"github.com/pierrec/lz4/v4"
+)
+
+func TestUntarLZ4(t *testing.T) {
+	for _, empty := range []bool{false, true} {
+		t.Run(fmt.Sprintf("empty=%t", empty), func(t *testing.T) {
+			want := map[string][]byte{}
+			if !empty {
+				want["small.txt"] = bytes.Repeat([]byte("small object\n"), 10)
+				want["large.txt"] = bytes.Repeat([]byte("large object\n"), 32768)
+			}
+			var compressed bytes.Buffer
+			compressor := lz4.NewWriter(&compressed)
+			archive := tar.NewWriter(compressor)
+			for name, data := range want {
+				if err := archive.WriteHeader(&tar.Header{Name: name, Mode: 0o600, Size: int64(len(data))}); err != nil {
+					t.Fatal(err)
+				}
+				if _, err := archive.Write(data); err != nil {
+					t.Fatal(err)
+				}
+			}
+			if err := archive.Close(); err != nil {
+				t.Fatal(err)
+			}
+			if err := compressor.Close(); err != nil {
+				t.Fatal(err)
+			}
+			var mu sync.Mutex
+			got := map[string][]byte{}
+			err := untar(t.Context(), iotest.HalfReader(bytes.NewReader(compressed.Bytes())), func(r io.Reader, info os.FileInfo, name string) error {
+				// Exercise a short first read followed by streaming the remaining content.
+				var output bytes.Buffer
+				prefix := make([]byte, 7)
+				if _, err := io.ReadFull(r, prefix); err != nil {
+					return err
+				}
+				output.Write(prefix)
+				if _, err := io.Copy(&output, r); err != nil {
+					return err
+				}
+				if int64(output.Len()) != info.Size() {
+					return fmt.Errorf("size mismatch for %s", name)
+				}
+				mu.Lock()
+				got[name] = output.Bytes()
+				mu.Unlock()
+				return nil
+			}, untarOptions{})
+			if err != nil {
+				t.Fatal(err)
+			}
+			if len(got) != len(want) {
+				t.Fatalf("objects=%d, want %d", len(got), len(want))
+			}
+			for name, data := range want {
+				if !bytes.Equal(got[name], data) {
+					t.Errorf("content mismatch for %s", name)
+				}
+			}
+			t.Run("corrupt-header", func(t *testing.T) {
+				damaged := bytes.Clone(compressed.Bytes())
+				damaged[6] ^= 0xff
+				err := untar(t.Context(), bytes.NewReader(damaged), func(io.Reader, os.FileInfo, string) error {
+					t.Error("corrupt header must be rejected before uploading objects")
+					return nil
+				}, untarOptions{})
+				if err == nil {
+					t.Fatal("corrupt LZ4 header accepted")
+				}
+			})
+		})
+	}
+}
@@ -0,0 +1,148 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"bytes"
+	"io"
+	"maps"
+	"net/http"
+	"testing"
+	"time"
+
+	"github.com/minio/minio/internal/bucket/replication"
+)
+
+func TestDeleteMarkerReadDataControls(t *testing.T) {
+	obj, er, disks, bucket := markerPurgeFixture(t, 4)
+	for _, body := range []string{"", "inline-data-control"} {
+		name := "data-" + mustGetUUID()
+		oi, err := er.PutObject(t.Context(), bucket, name, mustGetPutObjReader(t, bytes.NewBufferString(body), int64(len(body)), "", ""), ObjectOptions{Versioned: true})
+		if err != nil {
+			t.Fatal(err)
+		}
+		for _, disk := range disks {
+			fi, err := disk.ReadVersion(t.Context(), "", bucket, name, oi.VersionID, ReadOptions{ReadData: true})
+			if err != nil || fi.Deleted || fi.Size != int64(len(body)) || !fi.InlineData() || (body != "" && len(fi.Data) == 0) {
+				t.Fatalf("data read: size=%d inline=%v bytes=%d error=%v", fi.Size, fi.InlineData(), len(fi.Data), err)
+			}
+			// Legacy inline data without its annotation still gets the original
+			// ReadData behavior; the new marker guard must not affect it.
+			delete(fi.Metadata, ReservedMetadataPrefixLower+"inline-data")
+			if err := disk.WriteMetadata(t.Context(), "", bucket, name, fi); err != nil {
+				t.Fatal(err)
+			}
+			fi, err = disk.ReadVersion(t.Context(), "", bucket, name, oi.VersionID, ReadOptions{ReadData: true})
+			if err != nil || !fi.InlineData() {
+				t.Fatalf("legacy inline read: %+v error=%v", fi, err)
+			}
+		}
+		reader, err := obj.GetObjectNInfo(t.Context(), bucket, name, nil, http.Header{}, ObjectOptions{VersionID: oi.VersionID})
+		if err != nil {
+			t.Fatal(err)
+		}
+		got, err := io.ReadAll(reader)
+		reader.Close()
+		if err != nil || string(got) != body {
+			t.Fatalf("payload changed: %q error=%v", got, err)
+		}
+	}
+	_, opts := seedPurgeMarker(t, er, bucket, "stored-marker-key", false)
+	for _, disk := range disks {
+		fi, err := disk.ReadVersion(t.Context(), "", bucket, "stored-marker-key", opts.VersionID, ReadOptions{})
+		if err != nil {
+			t.Fatal(err)
+		}
+		// Preserve even an existing unusual marker key; suppress only the
+		// manufacture of a new inline annotation by the data-read branch.
+		fi.Metadata[ReservedMetadataPrefixLower+"inline-data"] = "original"
+		if err := disk.WriteMetadata(t.Context(), "", bucket, "stored-marker-key", fi); err != nil {
+			t.Fatal(err)
+		}
+		fi, err = disk.ReadVersion(t.Context(), "", bucket, "stored-marker-key", opts.VersionID, ReadOptions{ReadData: true})
+		if err != nil || fi.Metadata[ReservedMetadataPrefixLower+"inline-data"] != "original" {
+			t.Errorf("existing marker key changed: metadata=%v error=%v", fi.Metadata, err)
+		}
+	}
+}
+
+func TestDeleteMarkerMetadataRoundTrip(t *testing.T) {
+	stamp := time.Date(2026, 9, 1, 12, 1, 2, 345, time.UTC)
+	for _, free := range []bool{false, true} {
+		t.Run(map[bool]string{false: "replication", true: "free-version"}[free], func(t *testing.T) {
+			metadata := map[string]string{
+				ReservedMetadataPrefixLower + ReplicaStatus:        "REPLICA",
+				ReservedMetadataPrefixLower + ReplicaTimestamp:     stamp.Format(time.RFC3339Nano),
+				ReservedMetadataPrefixLower + ReplicationStatus:    "arn1=COMPLETED;arn2=FAILED;",
+				ReservedMetadataPrefixLower + ReplicationTimestamp: stamp.Add(-time.Hour).Format(time.RFC3339Nano),
+				VersionPurgeStatusKey:                              "arn1=PENDING;arn2=FAILED;",
+				targetResetHeader("arn1"):                          "original;reset1",
+				targetResetHeader("arn2"):                          "original;reset2",
+				ReservedMetadataPrefixLower + "unknown":            "",
+			}
+			if free {
+				metadata[ReservedMetadataPrefixLower+freeVersion] = ""
+				metadata[metaTierName] = "tier"
+				metadata[metaTierObjName] = "remote-object"
+				metadata[metaTierVersionID] = "remote-version"
+			}
+			fi := FileInfo{VersionID: mustGetUUID(), Deleted: true, ModTime: stamp, Metadata: maps.Clone(metadata)}
+			// Deliberately conflicting parsed state must never replace stored
+			// values or add a key missing from the raw metadata.
+			fi.ReplicationState = ReplicationState{ReplicaStatus: replication.Failed, ReplicaTimeStamp: stamp.Add(time.Hour), ResetStatusesMap: map[string]string{"new-arn": "invented"}}
+			fi.SetHealing()
+			fi.SetDataMov()
+			fi.SetTierFreeVersionID(mustGetUUID())
+			fi.SetTierFreeVersion()
+			fi.SetSkipTierFreeVersion()
+			xl := xlMetaV2{}
+			if err := xl.AddVersion(fi); err != nil {
+				t.Fatal(err)
+			}
+			stored, err := xl.getIdx(0)
+			if err != nil {
+				t.Fatal(err)
+			}
+			got := make(map[string]string)
+			for k, v := range stored.DeleteMarker.MetaSys {
+				got[k] = string(v)
+			}
+			if !maps.Equal(got, metadata) {
+				t.Errorf("stored metadata=%v want=%v", got, metadata)
+			}
+			decoded, err := xl.ToFileInfo("bucket", "marker", fi.VersionID, true, true)
+			if err != nil || !decoded.ModTime.Equal(fi.ModTime) || decoded.TierFreeVersion() != free {
+				t.Errorf("round trip identity: %+v error=%v", decoded, err)
+			}
+			if decoded.ReplicationState.ReplicaStatus != replication.Replica || decoded.ReplicationState.PurgeTargets["arn2"] != replication.VersionPurgeFailed || decoded.ReplicationState.Targets["arn1"] != replication.Completed {
+				t.Errorf("lost parsed replication state: %+v", decoded.ReplicationState)
+			}
+		})
+	}
+}
+
+func TestDeleteMarkerCreationMetadata(t *testing.T) {
+	_, er, disks, bucket := markerPurgeFixture(t, 4)
+	_, opts := seedPurgeMarker(t, er, bucket, "empty-key", false)
+	for i, disk := range disks {
+		fi, err := disk.ReadVersion(t.Context(), "", bucket, "empty-key", opts.VersionID, ReadOptions{})
+		if err != nil || fi.Metadata[ReservedMetadataPrefixLower+ReplicaStatus] != "REPLICA" || fi.Metadata[ReservedMetadataPrefixLower+ReplicaTimestamp] != opts.DeleteReplication.ReplicaTimeStamp.Format(time.RFC3339Nano) {
+			t.Errorf("disk %d lost new marker replica identity: metadata=%v error=%v", i, fi.Metadata, err)
+		}
+	}
+}
@@ -0,0 +1,67 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"strings"
+	"time"
+)
+
+// deleteMarkerMetadata preserves the complete stored marker state during
+// healing. The parsed ReplicationState is lossy; even absent raw keys must not
+// be reconstructed from it. Only creation writers without stored metadata
+// use typed state, and never invent timestamps.
+func deleteMarkerMetadata(fi FileInfo) map[string][]byte {
+	meta := make(map[string][]byte, len(fi.Metadata))
+	for k, v := range fi.Metadata {
+		switch k {
+		case xMinIOHealing, xMinIODataMov,
+			ReservedMetadataPrefixLower + tierFVID,
+			ReservedMetadataPrefixLower + tierFVMarker,
+			ReservedMetadataPrefixLower + tierSkipFVID:
+			continue
+		}
+		meta[k] = []byte(v)
+	}
+	if len(meta) != 0 || fi.Healing() || fi.DataMov() {
+		return meta
+	}
+	rs := fi.ReplicationState
+	if !rs.ReplicaStatus.Empty() {
+		meta[ReservedMetadataPrefixLower+ReplicaStatus] = []byte(rs.ReplicaStatus)
+		if !rs.ReplicaTimeStamp.IsZero() {
+			meta[ReservedMetadataPrefixLower+ReplicaTimestamp] = []byte(rs.ReplicaTimeStamp.UTC().Format(time.RFC3339Nano))
+		}
+	}
+	if rs.ReplicationStatusInternal != "" {
+		meta[ReservedMetadataPrefixLower+ReplicationStatus] = []byte(rs.ReplicationStatusInternal)
+		if !rs.ReplicationTimeStamp.IsZero() {
+			meta[ReservedMetadataPrefixLower+ReplicationTimestamp] = []byte(rs.ReplicationTimeStamp.UTC().Format(time.RFC3339Nano))
+		}
+	}
+	if rs.VersionPurgeStatusInternal != "" {
+		meta[VersionPurgeStatusKey] = []byte(rs.VersionPurgeStatusInternal)
+	}
+	for k, v := range rs.ResetStatusesMap {
+		if !strings.HasPrefix(k, ReservedMetadataPrefixLower+ReplicationReset) {
+			k = targetResetHeader(k)
+		}
+		meta[k] = []byte(v)
+	}
+	return meta
+}
@@ -0,0 +1,274 @@
+// Copyright (c) 2026 Feng Ruohang
+//
+// This file is part of Silo Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"fmt"
+	"math/rand"
+	"reflect"
+	"slices"
+	"testing"
+	"time"
+)
+
+// Use complete metadata so the selected version can also be decoded by LIST
+// and the object read path, rather than testing only synthetic shallow headers.
+func quorumNullVersion(t testing.TB, generation int64) xlMetaV2ShallowVersion {
+	t.Helper()
+	fi := newFileInfo("fixed/object", 12, 4)
+	fi.Erasure.Index = 1
+	fi.DataDir = "11111111-1111-1111-1111-111111111111"
+	fi.ModTime = time.Unix(generation, 0).UTC()
+	fi.Size = 8192
+	fi.Parts = []ObjectPartInfo{{Number: 1, Size: fi.Size, ActualSize: fi.Size}}
+	fi.Metadata = map[string]string{"etag": fmt.Sprintf("%032d", generation)}
+	var xl xlMetaV2
+	if err := xl.AddVersion(fi); err != nil {
+		t.Fatal(err)
+	}
+	return xl.versions[0]
+}
+
+// Start with forward, reverse and interleaved orders, then fixed-seed shuffles.
+func quorumVersionOrders(input [][]xlMetaV2ShallowVersion) [][][]xlMetaV2ShallowVersion {
+	orders := [][][]xlMetaV2ShallowVersion{slices.Clone(input), slices.Clone(input)}
+	slices.Reverse(orders[1])
+	interleaved := make([][]xlMetaV2ShallowVersion, 0, len(input))
+	for i, j := 0, len(input)-1; i <= j; i, j = i+1, j-1 {
+		interleaved = append(interleaved, input[i])
+		if i != j {
+			interleaved = append(interleaved, input[j])
+		}
+	}
+	orders = append(orders, interleaved)
+	for seed := range int64(32) {
+		order := slices.Clone(input)
+		rand.New(rand.NewSource(seed)).Shuffle(len(order), func(i, j int) {
+			order[i], order[j] = order[j], order[i]
+		})
+		orders = append(orders, order)
+	}
+	return orders
+}
+
+func TestMergeXLV2SingleNullQuorum(t *testing.T) {
+	v := []xlMetaV2ShallowVersion{quorumNullVersion(t, 100), quorumNullVersion(t, 200), quorumNullVersion(t, 300)}
+	for _, tt := range []struct {
+		name   string
+		counts [3]int
+		empty  int
+		quorum int
+		want   int
+	}{
+		{"consistent", [3]int{16}, 0, 8, 0},
+		{"old9-new7", [3]int{9, 7}, 0, 8, 0},
+		{"old15-new1", [3]int{15, 1}, 0, 8, 0},
+		{"old3-new1", [3]int{3, 1}, 0, 3, 0},
+		{"two-quorums", [3]int{8, 8}, 0, 8, 1},
+		{"empty-streams", [3]int{9, 1}, 6, 8, 0},
+		{"two-subquorums", [3]int{7, 7}, 2, 8, -1},
+		{"three-subquorums", [3]int{6, 5, 5}, 0, 8, -1},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			var input [][]xlMetaV2ShallowVersion
+			for g, n := range tt.counts {
+				for range n {
+					input = append(input, []xlMetaV2ShallowVersion{v[g]})
+				}
+			}
+			input = append(input, make([][]xlMetaV2ShallowVersion, tt.empty)...)
+			want := []xlMetaV2ShallowVersion{}
+			if tt.want >= 0 {
+				want = append(want, v[tt.want])
+			}
+			for order, versions := range quorumVersionOrders(input) {
+				for _, strict := range []bool{false, true} {
+					for _, requested := range []int{0, 1} {
+						got := mergeXLV2Versions(tt.quorum, strict, requested, versions...)
+						if !reflect.DeepEqual(got, want) {
+							t.Fatalf("order=%d strict=%v requested=%d: got %#v, want %#v", order, strict, requested, got, want)
+						}
+					}
+				}
+			}
+		})
+	}
+}
+
+func TestMergeXLV2SingleNullHeaderGroups(t *testing.T) {
+	old := quorumNullVersion(t, 100)
+	newer := quorumNullVersion(t, 200)
+	differentSig := old
+	differentSig.header.Signature[0]++
+	differentFlags := old
+	differentFlags.header.Flags ^= xlFlagInlineData
+	for _, tt := range []struct {
+		name   string
+		input  [][]xlMetaV2ShallowVersion
+		strict bool
+		want   []xlMetaV2ShallowVersion
+	}{
+		{"signature-nonstrict", [][]xlMetaV2ShallowVersion{{old}, {old}, {differentSig}, {newer}}, false, []xlMetaV2ShallowVersion{differentSig}},
+		{"signature-strict", [][]xlMetaV2ShallowVersion{{old}, {old}, {differentSig}, {newer}}, true, []xlMetaV2ShallowVersion{}},
+		{"flags-nonstrict", [][]xlMetaV2ShallowVersion{{old}, {old}, {differentFlags}, {newer}}, false, []xlMetaV2ShallowVersion{}},
+		{"flags-strict", [][]xlMetaV2ShallowVersion{{old}, {old}, {differentFlags}, {newer}}, true, []xlMetaV2ShallowVersion{}},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			got := mergeXLV2Versions(3, tt.strict, 0, tt.input...)
+			if !reflect.DeepEqual(got, tt.want) {
+				t.Fatalf("got %#v, want %#v", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestResolveSingleNullQuorum(t *testing.T) {
+	old, newer := quorumNullVersion(t, 100), quorumNullVersion(t, 200)
+	input := make([][]xlMetaV2ShallowVersion, 16)
+	for i := range input {
+		input[i] = []xlMetaV2ShallowVersion{old}
+		if i >= 9 {
+			input[i] = []xlMetaV2ShallowVersion{newer}
+		}
+	}
+	for order, versions := range quorumVersionOrders(input) {
+		entries := make(metaCacheEntries, len(versions))
+		for i := range versions {
+			xl := &xlMetaV2{versions: versions[i]}
+			metadata, err := xl.AppendTo(nil)
+			if err != nil {
+				t.Fatal(err)
+			}
+			entries[i] = metaCacheEntry{name: "fixed/object", metadata: metadata}
+		}
+		selected, ok := entries.resolve(&metadataResolutionParams{objQuorum: 8, dirQuorum: 8, requestedVersions: 1})
+		if !ok {
+			t.Fatalf("order %d: a complete null object version has quorum but was omitted", order)
+		}
+		xl, err := selected.xlmeta()
+		if err != nil {
+			t.Fatal(err)
+		}
+		if !reflect.DeepEqual(xl.versions, []xlMetaV2ShallowVersion{old}) {
+			t.Fatalf("order %d: incorrect complete selected version: %#v", order, xl.versions)
+		}
+		fi, err := xl.ToFileInfo("bucket", "fixed/object", "", false, false)
+		if err != nil || !fi.IsValid() || fi.Size != 8192 || !fi.ModTime.Equal(time.Unix(100, 0)) {
+			t.Fatalf("order %d: selected metadata cannot be read: %+v, %v", order, fi, err)
+		}
+	}
+}
+
+// These are compatibility assertions for inputs outside the narrow repair.
+// In particular, the mixed-version outputs below are not a new correctness
+// contract for version enumeration; that behavior requires a separate change.
+func TestMergeXLV2NullHistoriesUnchanged(t *testing.T) {
+	old, newer := quorumNullVersion(t, 100), quorumNullVersion(t, 300)
+	middle := quorumNullVersion(t, 200)
+	middle.header.VersionID = [16]byte{1}
+	highest := quorumNullVersion(t, 400)
+	highest.header.VersionID = [16]byte{2}
+	for _, tt := range []struct {
+		name  string
+		input [][]xlMetaV2ShallowVersion
+		want  []xlMetaV2ShallowVersion
+	}{
+		{"mixed-forward", [][]xlMetaV2ShallowVersion{{old}, {old}, {old}, {newer, middle}, {middle}, {middle}}, []xlMetaV2ShallowVersion{middle}},
+		{"mixed-reverse", [][]xlMetaV2ShallowVersion{{middle}, {middle}, {newer, middle}, {old}, {old}, {old}}, []xlMetaV2ShallowVersion{old}},
+		{"history-pruned-first", [][]xlMetaV2ShallowVersion{{highest, old}, {highest, old}, {highest, old}, {highest, newer}}, []xlMetaV2ShallowVersion{highest}},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			got := mergeXLV2Versions(3, false, 0, tt.input...)
+			if !reflect.DeepEqual(got, tt.want) {
+				t.Fatalf("excluded history changed: got %#v, want %#v", got, tt.want)
+			}
+		})
+	}
+	for _, kind := range []string{"delete", "free", "legacy", "mixed-ec", "legacy-modern-ec", "nonzero-strict"} {
+		t.Run(kind, func(t *testing.T) {
+			a, b := old, newer
+			switch kind {
+			case "delete":
+				b.header.Type = DeleteType
+			case "free":
+				b.header.Flags |= xlFlagFreeVersion
+			case "legacy":
+				b.header.Type = LegacyType
+			case "mixed-ec":
+				b.header.EcN, b.header.EcM = 8, 8
+			case "legacy-modern-ec":
+				b.header.EcN, b.header.EcM = 0, 0
+			case "nonzero-strict":
+				a.header.VersionID, b.header.VersionID = [16]byte{1}, [16]byte{1}
+			}
+			for _, requested := range []int{0, 1} {
+				got := mergeXLV2Versions(3, true, requested, []xlMetaV2ShallowVersion{a}, []xlMetaV2ShallowVersion{a}, []xlMetaV2ShallowVersion{a}, []xlMetaV2ShallowVersion{b})
+				if len(got) != 0 {
+					t.Fatalf("excluded input changed: requested=%d got %#v", requested, got)
+				}
+			}
+		})
+	}
+	// All-legacy EC headers are eligible, unlike mixing legacy and modern EC.
+	old.header.EcN, old.header.EcM = 0, 0
+	newer.header.EcN, newer.header.EcM = 0, 0
+	got := mergeXLV2Versions(3, false, 0, []xlMetaV2ShallowVersion{old}, []xlMetaV2ShallowVersion{old}, []xlMetaV2ShallowVersion{old}, []xlMetaV2ShallowVersion{newer})
+	if !reflect.DeepEqual(got, []xlMetaV2ShallowVersion{old}) {
+		t.Fatalf("all-legacy EC lost the old quorum: %#v", got)
+	}
+}
+
+func BenchmarkResolveSingleNullQuorumPage(b *testing.B) {
+	old, newer := quorumNullVersion(b, 100), quorumNullVersion(b, 200)
+	for _, kind := range []string{"consistent", "old-first", "new-first"} {
+		b.Run(kind, func(b *testing.B) {
+			var templates [16]metaCacheEntry
+			for i := range templates {
+				v := old
+				if kind == "old-first" && i >= 9 || kind == "new-first" && i < 7 {
+					v = newer
+				}
+				xl := &xlMetaV2{versions: []xlMetaV2ShallowVersion{v}}
+				data, err := xl.AppendTo(nil)
+				if err != nil {
+					b.Fatal(err)
+				}
+				templates[i] = metaCacheEntry{metadata: data}
+			}
+			var names [1000]string
+			for i := range names {
+				names[i] = fmt.Sprintf("fixed/%04d", i)
+			}
+			resolver := metadataResolutionParams{objQuorum: 8, dirQuorum: 8, requestedVersions: 1}
+			b.ReportAllocs()
+			b.ResetTimer()
+			for b.Loop() {
+				for _, name := range names {
+					entries := templates
+					for i := range entries {
+						entries[i].name = name
+					}
+					selected, ok := metaCacheEntries(entries[:]).resolve(&resolver)
+					if ok && selected.reusable {
+						metaDataPoolPut(selected.metadata)
+					}
+				}
+			}
+		})
+	}
+}
@@ -1630,7 +1630,7 @@ func (x *xlMetaV2) AddVersion(fi FileInfo) error {
 		ventry.DeleteMarker = &xlMetaV2DeleteMarker{
 			VersionID: uv,
 			ModTime:   fi.ModTime.UnixNano(),
-			MetaSys:   make(map[string][]byte),
+			MetaSys:   deleteMarkerMetadata(fi),
 		}
 	} else {
 		ventry.Type = ObjectType
@@ -1941,6 +1941,10 @@ func mergeXLV2Versions(quorum int, strict bool, requestedVersions int, versions
 		// No need for non-strict checks if quorum is 1.
 		strict = true
 	}
+	// Keep the original stream shapes: pruning must not make a versioned object
+	// eligible for the single null-version recount below.
+	originalVersions := versions
+	var checkedSingleNull, singleNull bool
 	// Shallow copy input
 	versions = append(make([][]xlMetaV2ShallowVersion, 0, len(versions)), versions...)

@@ -2015,44 +2019,23 @@ func mergeXLV2Versions(quorum int, strict bool, requestedVersions int, versions
 					// Version IDs match, but otherwise unable to resolve.
 					// We are either strict, or don't have enough information to match.
 					// Switch to a pure counting algo.
-					x := make(map[xlMetaV2VersionHeader]int, len(tops))
-					for _, a := range tops {
-						if a.header.VersionID != ver.header.VersionID {
-							continue
-						}
-						if !strict {
-							// we must match EC, when we are not strict.
-							if !a.header.matchesEC(ver.header) {
-								continue
-							}
-
-							a.header.Signature = [4]byte{}
-						}
-						x[a.header]++
-					}
-					latestCount = 0
-					for k, v := range x {
-						if v < latestCount {
-							continue
-						}
-						if v == latestCount && latest.header.sortsBefore(k) {
-							// Tiebreak, use sort.
-							continue
-						}
-						for _, a := range tops {
-							hdr := a.header
-							if !strict {
-								hdr.Signature = [4]byte{}
-							}
-							if hdr == k {
-								latest = a
-							}
-						}
-						latestCount = v
-					}
+					latest, latestCount = countXLV2Versions(tops, ver.header, latest, strict)
 					break
 				}
 			}
+			if latestCount < quorum {
+				if !checkedSingleNull {
+					singleNull = singleNullVersionStreams(originalVersions)
+					checkedSingleNull = true
+				}
+				if singleNull {
+					// A newer minority at the end can hide an older quorum from
+					// the selection loop. Recount before discarding the null ID.
+					if candidate, count := countXLV2Versions(tops, latest.header, latest, strict); count >= quorum {
+						latest, latestCount = candidate, count
+					}
+				}
+			}
 			if latestCount >= quorum {
 				merged = append(merged, latest)

@@ -2113,6 +2096,70 @@ func mergeXLV2Versions(quorum int, strict bool, requestedVersions int, versions
 	return merged
 }

+// singleNullVersionStreams excludes histories and other version types from the
+// additional recount. Check the original inputs, before any stream is pruned.
+func singleNullVersionStreams(versions [][]xlMetaV2ShallowVersion) bool {
+	var ec xlMetaV2VersionHeader
+	var haveEC bool
+	for _, stream := range versions {
+		if len(stream) == 0 {
+			continue
+		}
+		if len(stream) != 1 {
+			return false
+		}
+		h := stream[0].header
+		if h.VersionID != [16]byte{} || h.Type != ObjectType || h.FreeVersion() {
+			return false
+		}
+		if haveEC && (h.EcN != ec.EcN || h.EcM != ec.EcM) {
+			return false
+		}
+		ec, haveEC = h, true
+	}
+	return haveEC
+}
+
+// countXLV2Versions selects the most frequent compatible header for reference's
+// VersionID, retaining the existing sort tiebreak and last matching entry.
+func countXLV2Versions(tops []xlMetaV2ShallowVersion, reference xlMetaV2VersionHeader, latest xlMetaV2ShallowVersion, strict bool) (xlMetaV2ShallowVersion, int) {
+	x := make(map[xlMetaV2VersionHeader]int, len(tops))
+	for _, a := range tops {
+		if a.header.VersionID != reference.VersionID {
+			continue
+		}
+		if !strict {
+			// we must match EC, when we are not strict.
+			if !a.header.matchesEC(reference) {
+				continue
+			}
+			a.header.Signature = [4]byte{}
+		}
+		x[a.header]++
+	}
+	var latestCount int
+	for k, v := range x {
+		if v < latestCount {
+			continue
+		}
+		if v == latestCount && latest.header.sortsBefore(k) {
+			// Tiebreak, use sort.
+			continue
+		}
+		for _, a := range tops {
+			hdr := a.header
+			if !strict {
+				hdr.Signature = [4]byte{}
+			}
+			if hdr == k {
+				latest = a
+			}
+		}
+		latestCount = v
+	}
+	return latest, latestCount
+}
+
 type xlMetaBuf []byte

 // ToFileInfo converts xlMetaV2 into a common FileInfo datastructure
@@ -1719,7 +1719,9 @@ func (s *xlStorage) ReadVersion(ctx context.Context, origvolume, volume, path, v
 		defer metaDataPoolPut(buf)
 	}

-	if readData {
+	// Delete markers have no payload. In particular, do not manufacture an
+	// inline-data metadata key when healing their zero-size FileInfo.
+	if readData && !fi.Deleted {
 		if len(fi.Data) > 0 || fi.Size == 0 {
 			if fi.InlineData() {
 				// If written with header we are fine.
@@ -2,7 +2,7 @@

 PR #60 introduced an opt-in scheduler that moved objects between local server pools according to GET frequency. It and its feature-specific fixes have been removed. This does not remove ordinary lifecycle expiration, transitions to remote tiers, rebalance, decommission, or the general multi-pool correctness fixes from PR #178.

-The [introduction and rollback record](../../investigations/access-tiering-revert.md) documents the commit history, scope decision, review corrections and unresolved validation findings.
+The [introduction and rollback record](https://silo.pgsty.com/compatibility/access-tiering-removal/) documents the commit history, scope decision, review corrections and unresolved validation findings.

 The published Server 20260903 predates this feature. These instructions concern main/snapshot deployments that included #60; upgrading from the published version does not require access-tier configuration cleanup.

@@ -35,7 +35,7 @@ set -e
 export MINIO_CI_CD=1
 export MINIO_BROWSER=off
 export MINIO_ROOT_USER="minio"
-export MINIO_ROOT_PASSWORD="silo123"
+export MINIO_ROOT_PASSWORD="silo12345"
 export MINIO_KMS_AUTO_ENCRYPTION=off
 export MINIO_PROMETHEUS_AUTH_TYPE=public
 export MINIO_KMS_SECRET_KEY=my-minio-key:OSMM+vkKUTCvQs9YL/CVMIMt43HFhkUpqJxTmGl6rYw=
@@ -58,8 +58,8 @@ silo server --address 127.0.0.1:9003 "http://127.0.0.1:9003/tmp/multisiteb/data/
 silo server --address 127.0.0.1:9004 "http://127.0.0.1:9003/tmp/multisiteb/data/disterasure/xl{1...4}" \
 	"http://127.0.0.1:9004/tmp/multisiteb/data/disterasure/xl{5...8}" >/tmp/siteb_2.log 2>&1 &

-export MC_HOST_sitea=http://minio:silo123@127.0.0.1:9001
-export MC_HOST_siteb=http://minio:silo123@127.0.0.1:9004
+export MC_HOST_sitea=http://minio:silo12345@127.0.0.1:9001
+export MC_HOST_siteb=http://minio:silo12345@127.0.0.1:9004

 ./mc ready sitea
 ./mc ready siteb
@@ -83,7 +83,7 @@ done

 echo "adding replication rule for site a -> site b"
 ./mc replicate add sitea/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/bucket
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/bucket

 remote_arn=$(./mc replicate ls sitea/bucket --json | jq -r .rule.Destination.Bucket)
 sleep 1
@@ -137,7 +137,7 @@ fi

 echo "adding replication rule for site a -> site b"
 ./mc replicate add sitea/bucket-version/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/bucket-version
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/bucket-version

 ./mc mb sitea/bucket-version/directory/

@@ -37,7 +37,7 @@ set -e
 export MINIO_CI_CD=1
 export MINIO_BROWSER=off
 export MINIO_ROOT_USER="minio"
-export MINIO_ROOT_PASSWORD="silo123"
+export MINIO_ROOT_PASSWORD="silo12345"
 export MINIO_KMS_AUTO_ENCRYPTION=off
 export MINIO_PROMETHEUS_AUTH_TYPE=public
 export MINIO_KMS_SECRET_KEY=my-minio-key:OSMM+vkKUTCvQs9YL/CVMIMt43HFhkUpqJxTmGl6rYw=
@@ -70,9 +70,9 @@ silo server --address 127.0.0.1:9005 "http://127.0.0.1:9005/tmp/multisitec/data/
 silo server --address 127.0.0.1:9006 "http://127.0.0.1:9005/tmp/multisitec/data/disterasure/xl{1...4}" \
 	"http://127.0.0.1:9006/tmp/multisitec/data/disterasure/xl{5...8}" >/tmp/sitec_2.log 2>&1 &

-export MC_HOST_sitea=http://minio:silo123@127.0.0.1:9001
-export MC_HOST_siteb=http://minio:silo123@127.0.0.1:9004
-export MC_HOST_sitec=http://minio:silo123@127.0.0.1:9006
+export MC_HOST_sitea=http://minio:silo12345@127.0.0.1:9001
+export MC_HOST_siteb=http://minio:silo12345@127.0.0.1:9004
+export MC_HOST_sitec=http://minio:silo12345@127.0.0.1:9006

 ./mc ready sitea
 ./mc ready siteb
@@ -93,73 +93,73 @@ export MC_HOST_sitec=http://minio:silo123@127.0.0.1:9006
 echo "adding replication rule for a -> b : ${remote_arn}"
 sleep 1
 ./mc replicate add sitea/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync"
 sleep 1

 echo "adding replication rule for b -> a : ${remote_arn}"
 ./mc replicate add siteb/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9001/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9001/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync"
 sleep 1

 echo "adding replication rule for a -> c : ${remote_arn}"
 ./mc replicate add sitea/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9006/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9006/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 2
 sleep 1

 echo "adding replication rule for c -> a : ${remote_arn}"
 ./mc replicate add sitec/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9001/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9001/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 2
 sleep 1

 echo "adding replication rule for b -> c : ${remote_arn}"
 ./mc replicate add siteb/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9006/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9006/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 3
 sleep 1

 echo "adding replication rule for c -> b : ${remote_arn}"
 ./mc replicate add sitec/bucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/bucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/bucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 3
 sleep 1

 echo "adding replication rule for olockbucket a -> b : ${remote_arn}"
 ./mc replicate add sitea/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync"
 sleep 1

 echo "adding replication rule for olockbucket b -> a : ${remote_arn}"
 ./mc replicate add siteb/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9001/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9001/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync"
 sleep 1

 echo "adding replication rule for olockbucket a -> c : ${remote_arn}"
 ./mc replicate add sitea/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9006/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9006/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 2
 sleep 1

 echo "adding replication rule for olockbucket c -> a : ${remote_arn}"
 ./mc replicate add sitec/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9001/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9001/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 2
 sleep 1

 echo "adding replication rule for olockbucket b -> c : ${remote_arn}"
 ./mc replicate add siteb/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9006/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9006/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 3
 sleep 1

 echo "adding replication rule for olockbucket c -> b : ${remote_arn}"
 ./mc replicate add sitec/olockbucket/ \
-	--remote-bucket http://minio:silo123@127.0.0.1:9004/olockbucket \
+	--remote-bucket http://minio:silo12345@127.0.0.1:9004/olockbucket \
 	--replicate "existing-objects,delete,delete-marker,replica-metadata-sync" --priority 3
 sleep 1

@@ -204,33 +204,33 @@ head -c 221227088 </dev/urandom >200M
 sleep 10

 echo "Verifying ETag for all objects"
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9001/ -bucket bucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9002/ -bucket bucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9003/ -bucket bucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9004/ -bucket bucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9005/ -bucket bucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9006/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9001/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9002/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9003/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9004/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9005/ -bucket bucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9006/ -bucket bucket

-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9001/ -bucket olockbucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9002/ -bucket olockbucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9003/ -bucket olockbucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9004/ -bucket olockbucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9005/ -bucket olockbucket
-./s3-check-md5 -versions -access-key minio -secret-key silo123 -endpoint http://127.0.0.1:9006/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9001/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9002/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9003/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9004/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9005/ -bucket olockbucket
+./s3-check-md5 -versions -access-key minio -secret-key silo12345 -endpoint http://127.0.0.1:9006/ -bucket olockbucket

 # additional tests for encryption object alignment
 go install -v github.com/minio/multipart-debug@latest

-upload_id=$(multipart-debug --endpoint 127.0.0.1:9001 --accesskey minio --secretkey silo123 multipart new --bucket bucket --object new-test-encrypted-object --encrypt)
+upload_id=$(multipart-debug --endpoint 127.0.0.1:9001 --accesskey minio --secretkey silo12345 multipart new --bucket bucket --object new-test-encrypted-object --encrypt)

 dd if=/dev/urandom bs=1 count=7048531 of=/tmp/7048531.txt
 dd if=/dev/urandom bs=1 count=2847391 of=/tmp/2847391.txt

 sudo apt install jq -y

-etag_1=$(multipart-debug --endpoint 127.0.0.1:9002 --accesskey minio --secretkey silo123 multipart upload --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} --file /tmp/7048531.txt --number 1 | jq -r .ETag)
-etag_2=$(multipart-debug --endpoint 127.0.0.1:9001 --accesskey minio --secretkey silo123 multipart upload --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} --file /tmp/2847391.txt --number 2 | jq -r .ETag)
-multipart-debug --endpoint 127.0.0.1:9002 --accesskey minio --secretkey silo123 multipart complete --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} 1.${etag_1} 2.${etag_2}
+etag_1=$(multipart-debug --endpoint 127.0.0.1:9002 --accesskey minio --secretkey silo12345 multipart upload --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} --file /tmp/7048531.txt --number 1 | jq -r .ETag)
+etag_2=$(multipart-debug --endpoint 127.0.0.1:9001 --accesskey minio --secretkey silo12345 multipart upload --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} --file /tmp/2847391.txt --number 2 | jq -r .ETag)
+multipart-debug --endpoint 127.0.0.1:9002 --accesskey minio --secretkey silo12345 multipart complete --bucket bucket --object new-test-encrypted-object --uploadid ${upload_id} 1.${etag_1} 2.${etag_2}

 sleep 10

@@ -35,7 +35,7 @@ set -e
 export MINIO_CI_CD=1
 export MINIO_BROWSER=off
 export MINIO_ROOT_USER="minio"
-export MINIO_ROOT_PASSWORD="silo123"
+export MINIO_ROOT_PASSWORD="silo12345"
 export MINIO_KMS_AUTO_ENCRYPTION=off
 export MINIO_PROMETHEUS_AUTH_TYPE=public
 export MINIO_KMS_SECRET_KEY=my-minio-key:OSMM+vkKUTCvQs9YL/CVMIMt43HFhkUpqJxTmGl6rYw=
@@ -70,10 +70,10 @@ silo server --address 127.0.0.1:9008 "http://127.0.0.1:9007/tmp/multisited/data/

 # Wait to make sure all Silo instances are up

-export MC_HOST_sitea=http://minio:silo123@127.0.0.1:9001
-export MC_HOST_siteb=http://minio:silo123@127.0.0.1:9004
-export MC_HOST_sitec=http://minio:silo123@127.0.0.1:9006
-export MC_HOST_sited=http://minio:silo123@127.0.0.1:9008
+export MC_HOST_sitea=http://minio:silo12345@127.0.0.1:9001
+export MC_HOST_siteb=http://minio:silo12345@127.0.0.1:9004
+export MC_HOST_sitec=http://minio:silo12345@127.0.0.1:9006
+export MC_HOST_sited=http://minio:silo12345@127.0.0.1:9008

 ./mc ready sitea
 ./mc ready siteb
@@ -89,7 +89,7 @@ export MC_HOST_sited=http://minio:silo123@127.0.0.1:9008
 sleep 10s

 ## Add warm tier
-./mc ilm tier add minio sitea WARM-TIER --endpoint http://localhost:9006 --access-key minio --secret-key silo123 --bucket bucket
+./mc ilm tier add minio sitea WARM-TIER --endpoint http://localhost:9006 --access-key minio --secret-key silo12345 --bucket bucket

 ## Add ILM rules
 ./mc ilm add sitea/bucket --transition-days 0 --transition-tier WARM-TIER --transition-days 0 --noncurrent-expire-days 2 --expire-days 3 --prefix "myprefix" --tags "tag1=val1&tag2=val2"
@@ -1,5 +1,7 @@
 # Compression Guide

+For SILO's current SSE-C compression and historical-object boundary, see the [maintained encryption guide](https://silo.pgsty.com/administration/server-side-encryption/server-side-encryption-sse-c/) and [SSE-C replica design](https://silo.pgsty.com/blog/design/ssec-replica-integrity/). Check the documented release boundary before applying main-branch behavior to an older binary.
+
 Silo server allows streaming compression to ensure efficient disk space usage.
 Compression happens inflight, i.e objects are compressed before being written to disk(s).
 Silo uses [`klauspost/compress/s2`](https://github.com/klauspost/compress/tree/master/s2)
@@ -1,5 +1,7 @@
 # Federation Quickstart Guide *Federation feature is deprecated and should be avoided for future deployments*

+The maintained [federated CopyObject design](https://silo.pgsty.com/blog/design/federated-copy-object/) records the destination encryption, checksum, Object Lock and committed-response contract, including its Server release boundary.
+
 This document explains how to configure Silo with `Bucket lookup from DNS` style federation.

 ## Cross-deployment copy behavior
@@ -38,7 +38,7 @@ pid=$!

 mc ready mysilo

-mc admin user add mysilo/ silo123 silo123
+mc admin user add mysilo/ silo123 silo12345

 mc admin policy create mysilo/ deny-non-sse-kms-pol ./docs/iam/policies/deny-non-sse-kms-objects.json
 mc admin policy create mysilo/ deny-invalid-sse-kms-pol ./docs/iam/policies/deny-objects-with-invalid-sse-kms-key-id.json
@@ -50,7 +50,7 @@ mc admin policy attach mysilo consoleAdmin --user silo123
 mc mb -l mysilo/test-bucket
 mc mb -l mysilo/multi-key-poc

-export MC_HOST_mysilo1="http://silo123:silo123@localhost:9000/"
+export MC_HOST_mysilo1="http://silo123:silo12345@localhost:9000/"

 mc cp /etc/issue mysilo1/test-bucket
 ret=$?
@@ -1,193 +0,0 @@
-# 访问频率分层：引入、修复与回退记录
-
-记录日期：2026-09-15。本记录说明 PR #60 的设计、合入后的取舍，以及此次回退为什么同时保留并补齐通用多池正确性修复。操作步骤见[退役迁移说明](../bucket/lifecycle/access-tiering-removal.md)。合并、正式发布和生产部署是不同状态；本记录随回退变更交付，不代表已经发布。
-
-初稿审查时（2026-09-15 03:20 UTC），最终候选尚未移植到远端基线、尚未冻结 PR head，合并前全量检查与三次有效 Linux 运行尚未执行，本变更尚未合并。后续执行状态以承载本记录的 PR 及其绑定提交的验收记录为准；下文的历史实验不替代这些检查。
-
-**当前验收状态：** 回退、普通版本 DELETE 调和、扫描复用及文档由 [PR #188](https://github.com/pgsty/silo/pull/188) 交付。本地与 CI 检查通过；首次 Linux 验收因 DELETE204 后的 HEAD/GET503 停止。随后完成可控机制实验、匹配写入负载对照，并修正准备检查；独立的新轮次 R-Upgrade-2 三次完整升级验收均通过。旧失败没有改判，原单次请求的逐盘状态仍不可追溯。最终合并状态以 PR 为准，正式发布与部署另行验收。详情见第 9 至 11 节。
-
-## 1. 引入的目标和实际范围
-
-[@mrjavadseydi](https://github.com/mrjavadseydi) 在 [PR #60](https://github.com/pgsty/silo/pull/60) 提出了基于 GET 频率的本地池间分层。成功 GET 更新有界滚动计数，后台调度器把热对象提升到配置中的首个池，把已经迁移且变冷的对象降到末个池。功能默认关闭，至少需要两个池；它与普通生命周期过期、远端对象存储 transition、rebalance 和 decommission 是不同机制。
-
-实现不只是一个后台任务：它增加了 GET 计数入口、leader 调度、跨池版本栈复制、来源清理及失败恢复、热池配额、生命周期 XML 扩展、配置项、指标和扫描统计。访问热度统计使 data-usage cache 从 v8 升到 v9。对象本体的存储格式没有因此改变。
-
-搬移要保持完整版本历史、删除标记、null version、时间戳、ETag、校验和和加密元数据；同时需要处理并发写入、目的端已有版本、来源部分删除和远端 tier 引用。合入前的修补和测试针对并覆盖了这些边界，不应把回退解释成贡献无效。贡献者署名继续保留，独立的其他贡献也不回退。
-
-## 2. 可追溯时间线
-
-下表的 PR/Issue 时间使用 UTC；提交链接对应具体代码，不把“报告时间”当作“缺陷首次出现时间”。
-
-| 时间 | 事件 | 本次处理 |
-| --- | --- | --- |
-| 2026-08-15 10:15 | #60 创建；原始实现 [`7a060cab1`](https://github.com/pgsty/silo/commit/7a060cab1edd5bbc17da7f703bbd1ab7415b6f7c) | 随特性撤销 |
-| 2026-09-06 00:09 | [#133](https://github.com/pgsty/silo/issues/133) 报告多池副本写不能权威调和 Object Lock 状态 | 保留解决它的通用修复 |
-| 2026-09-06 15:12 | [#144](https://github.com/pgsty/silo/issues/144) 报告条件 DELETE 原子性仅限单个纠删码集合 | 保留解决它的通用修复 |
-| 2026-09-08 05:18 | [`9a6e1477f`](https://github.com/pgsty/silo/commit/9a6e1477f45067559def8423d431ee177795134f) 补访问分层兼容标识清单 | 删除功能专属标识，保留有依据的退役兼容 |
-| 2026-09-08 07:08 | [`374de0fa3`](https://github.com/pgsty/silo/commit/374de0fa32aa1d6eda57dbba6ac522ebf793b6be) 修复搬移的版本保全、写隔离和删除范围 | 随专属搬移器撤销 |
-| 2026-09-08 07:28 | [`5ac33e158`](https://github.com/pgsty/silo/commit/5ac33e1583e838ade3f56c7d60e80e7a854f9a88) 确定性覆盖搬移失败恢复 | 随已删除搬移器的专属测试撤销 |
-| 2026-09-08 07:42 | #60 以 [`a3df317ae`](https://github.com/pgsty/silo/commit/a3df317ae0725eb650e4d3e21551154f69be6229) 合入 | 以该 merge 的第一父差异确定功能边界 |
-| 2026-09-11 12:34 | [#178](https://github.com/pgsty/silo/pull/178) 合入通用多池写入、元数据与条件删除调和 | 保留，解除测试对访问搬移器的依赖 |
-| 2026-09-13 | [`2dd1e00da`](https://github.com/pgsty/silo/commit/2dd1e00da49faf995f2db29807fc6211b8376d7d) 的 CHANGELOG 同时汇总访问分层和通用多池修复 | 拆开表述，不整条删除独立修复历史 |
-| 2026-09-15 | 维护者决定收缩访问分层；完成来源分析、三种候选反证、退役兼容、普通版本 DELETE 修补与外部评审 | 形成此次选择性回退 |
-
-#178 中的 [`e59a3d938`](https://github.com/pgsty/silo/commit/e59a3d938ed25c1bcd51efbb4ad6955073d195f7) 提供池级串行化、字段调和和条件删除；[`ccb676e60`](https://github.com/pgsty/silo/commit/ccb676e60cb7441ee65ff7c35f3b7828979101fd) 保护仍被其他副本使用的远端 tier 引用；[`51d41345f`](https://github.com/pgsty/silo/commit/51d41345f7ac532f9c6fea2b7dba5f29da930b8f) 保存 Linux 重启及 OIDC 验收记录。这三项不属于仅为访问频率调度而存在的代码。
-
-截至回退评估时，公开 Server `RELEASE.2026-09-03T13-18-01Z` 早于 #60 合入。退役迁移主要针对运行过后续 main、自行构建或快照版本的实例；不能据此声称正式 release 用户普遍启用过此特性。
-
-## 3. 为什么回退，为什么不能整批撤销后续修复
-
-维护者的取舍是：默认关闭的可选调度能力，对配置、生命周期、缓存、统计和核心多池写入路径带来了过大的维护面。此次移除的是该能力及专属实现，没有测量并宣称吞吐量提升、延迟降低或固定减少一次分布式锁往返。
-
-“后来改过同一文件”不等于“由 #60 引发”。#133/#144 的报告早于 #60 合入；更关键的是，原有 rebalance、decommission 和复制写入也会使同一版本暂时存在于多个池。访问分层消失后，这些状态仍然合法存在。移除调和与锁纪律会重新允许旧副本遮蔽新元数据、条件删除选错版本、清理错误被吞掉等问题。
-
-评估在隔离工作树中实际比较了三条路线：
-
-| 候选 | 通用多池回归 | 普通指定版本 DELETE |
-| --- | --- | --- |
-| A：撤销 #60，保留 #178 | 原有 13 组通过 | 仍能成功返回后留下可读副本 |
-| B：同时撤销 #60 和 #178 的存储修补 | 相同 13 组中 10 组失败 | 问题仍在 |
-| C：A 加普通版本 DELETE 调和 | 13 组原有及当时新增的 8 组通过 | 同一复现通过 |
-
-这些是 2026-09-15 的历史对照结果，不是最终 PR head 的发布验收。后续补上目录标记、真实 rebalance 中断等覆盖后，通用多池测试达到 23 组。测试通过不能替代来源分析，来源分析也不能替代最终候选的运行验证。
-
-## 4. 最终保留与删除的边界
-
- 删除访问 tracker、调度/搬移器、GET 和 scanner 钩子、热池配额、专属配置帮助、生命周期动作、指标及专属测试。
- 保留普通过期、远端 transition、rebalance/decommission、复制写入，以及 #178 的 Object Lock、标签、条件删除、元数据调和与远端引用保护。
- 原有 data-usage 生成解码器恢复到 #60 前的实现；允许读取 v8/v9，利用字段编码跳过已退役热度字段，继续写 v8。普通字段保真由历史真实 v9 样本测试覆盖。
- 仅容忍准确的十个退役 ILM 键；读取生命周期时丢弃退役扩展。纯访问动作的规则需先清理才能再次编辑；混合规则保留普通动作。
- 已经搬移的对象留在当前池；没有全量搬回、自动删除所有重复版本、后台清理服务或对象元数据重写。
-
-曾对本地审查提交 `d06f1c614` 做过声明来源复核：消失的 357 个声明均不在 #60 之前，删除的十个文件均由 #60 引入；#60 原先删换的 41 行旧文本按忽略空白比较有 40 行恢复，剩下一行保留 #178 在已持有池锁时调用 `getWritePoolIdx(..., true)` 的修正，避免对同一对象再次取锁。生成缓存解码器与功能前逐字节一致，原有 13 组通用测试没有删除。这是该审查版本的保全证据，不能把数字脱离 SHA 当作未来所有版本的保证。
-
-## 5. 普通版本 DELETE 是独立补洞
-
-功能移除不会自动消除历史重复副本。普通单对象指定版本 DELETE 因而复用既有调和路径：在池级对象锁内读取每个池的目标版本，计算一次条件及回调，先删除非权威副本，再处理权威副本；任何不可读池或清理错误都不能当作成功。
-
-范围包括 UUID、null version、delete marker，以及原先就被解析为 null version 的未指定版本目录标记 DELETE。入站复制、搬移内部调用、生命周期过期和 free-version 清理保留各自语义；批量 `DeleteObjects` 原本就会向池并发扇出，不是此次遗漏。
-
-删除标记需要向 retention/metadata 回调传入与 set 层相同的 `MethodNotAllowed` 或 `ObjectNotFound` 语义。直接复用拒绝 marker 的元数据更新入口会错误地拒绝合法版本删除。回调从所有副本合并独立更新的 Object Lock 和标签，不能随意只采用一个池的状态。
-
-存在两项明确的成功/失败边界：
-
-1. 读法定多数不足时返回 `503 SlowDownRead`，即使另一个池有可读副本。旧路径的结果会受池遍历顺序影响；新路径把失败语义统一。它是正确性与可用性的取舍，需要恢复后重试。
-2. 出站删除复制尚未完成时，成功响应可以表示各副本进入 `VersionPurgePending`，由既有 worker 完成清理。原有每池 quorum 规则也继续适用；不能把成功响应等同于每一块盘立即物理删除。
-
-## 6. 审查如何改变了方案
-
-本地先形成三个线性审查提交：`8fdfdabd9` 移除特性，`d06f1c614` 补普通 DELETE，`6e3fdca97` 去除重复扫描。它们记录审查演进，最终 PR 在独立远端基线上重放，提交 ID 会改变；不应把线性演进误认为三份同时维护的实现。
-
-Claude Code Opus 5 / max 的五轮实现评审要求补齐退役兼容、说明协调停机及环境一致性、验证真实池故障，并纠正运行证据措辞。随后 Claude 与 ZCode 的独立复核再次确认了回退边界和普通 DELETE 语义；最终两轮计划商榷收束了交付流程。
-
-| 意见 | 裁定与处理 |
-| --- | --- |
-| 指定版本 DELETE 重复扫描所有池 | 接受。第一次扫描已在同一锁内得到目标副本，直接合并其结果。16 盘池在回调前的读取计数从 32 降为 16；这不等于总 I/O 或延迟减半。 |
-| N 个副本产生 N 条 DELETE 审计 | 反驳。底层调用只追加上下文标签，HTTP 层向每个配置审计目标发送一次请求事件。成功完成调和时池标签最终指向 primary；不增加额外 NoAuditLog 修改。 |
-| retention/metadata 顺序与 set 层不同 | 差异存在，但已有明确注释。保留 retention 优先，拒绝后不删除、不调度删除复制、不 Sweep；不宣称任意自定义回调都能交换。 |
-| 把优化 amend 进旧提交并直接丢弃 main 脏修改 | 不 amend 已审历史。先保存完整文件、补丁、哈希及可达恢复引用，核对覆盖后受控恢复。八组新增通用测试承接，访问搬移专用测试由真实 rebalance 中断覆盖替代。 |
-| 从当前本地分支直接开 PR | 调整。其祖先含另一个任务的 IAM/超时提交 `ebc9937d9`；从远端 `89637554d` 仅移植本次变更，不夹带或删除独立工作。 |
-| 合并之后再跑全量与 Linux 验收 | 不接受。先完成文档和移植，冻结 PR head，再验收；不能把旧 SHA 的测试直接提升为新基线的通过记录。 |
-| 不同基线 diff 必须逐字节一致 | 改为每提交 stable patch-id、路径与完整树等价性核验。blob hash 和行号随基线变化，不应成为错误的拒绝依据。 |
-
-`objectPoolInfos` 的并行查询作为独立性能跟进，本次不增加并发实现。Contributor 署名、#132 配额指标、#77 桶元数据、federated COPY、IAM、超时及其他独立修复不因本次取舍被整体撤销。
-
-## 7. 历史运行证据与未定案事项
-
-以下是移植前 `d06f1c614` 的历史实验，不是最终 PR head 的验收替身。
-
-| 运行 | 实际观察 | 不应推导的结论 |
-| --- | --- | --- |
-| `f590538f`，四节点双池 | 旧版真实 rebalance 中断留下 9 个重复 UUID；协调停机换新后对象四节点可读，旧配置与普通规则可编辑，真实缓存头 v9→v8；删除一个 addressed UUID 后四节点 HEAD/GET 404，另一个版本仍可读 | 没有验证全部 9 个不同 UUID 收敛；物理元数据仅每池抽查一盘；没有整份运行时统计守恒测量 |
-| `fd93cd37`，三节点双池 | 停掉一个池所在节点，保留 namespace 锁的 2/3 法定多数；DELETE 返回 503 SlowDownRead，源全部四盘保留目标版本；恢复后 DELETE204，各节点 HEAD/GET404 | 不能推广为跨主机网络、持久盘及压力验收 |
-| 被弃用的四节点停池拓扑 | 同时丢失 namespace 锁法定多数，发生客户端超时，记录脚本还遇到 NoneType 错误 | 不是“池读取返回503”的证明 |
-| `83676ca2` | 升级后配置/生命周期编辑之后一次 HeadObject 返回503；artifact 没有记录 DELETE 自身响应 | “DELETE之后”仅来自脚本顺序，不能写成已证实 DELETE204 后异常，也不能归类为已修复、既有问题或暂态 |
-
-**开放项：`83676ca2` 的 HEAD503 仍未定案。** 可直接比较的目标阶段历史运行是一失败、一成功；一次未复现不足以关闭问题。仅凭 `SlowDownWrite` 等错误名字也不能给其他失败确定容量或环境根因。
-
-最终候选的合并前复核采用有界规则：要求三次有效升级后 DELETE/HEAD 运行，最多五次总尝试；每次记录 DELETE 码/耗时、失败 HEAD 的节点与版本、GET 错误码、两池全盘元数据、固定间隔重试时序和旧版同拓扑对照。旧版可能保留副本返回200，不要求它满足新增跨池删除契约。
-
-有证据证明在目标操作前失败的 harness 尝试才能不计入有效运行，但仍计入总尝试。任何目标阶段的新503或数据不变量失败都不能通过补跑抹掉，必须暂停合并并定位。三次通过也只满足这项工程检查，不证明历史异常已消失；开放项在合并和正式发布评估时仍须可见。
-
-## 8. 交付与恢复纪律
-
-最终候选使用独立分支；main 的五个旧修改先保存完整内容、二进制补丁、SHA-256 和具名 Git 恢复引用，再核对原 HEAD/哈希及测试覆盖，只恢复这五个文件。禁止用整树 reset 或 clean 代替受控归一；若用户已有新增编辑，应保留并重新核对。
-
-全量 cmd/internal、相关 race、构建、vet、lint、生成文件和兼容检查，以及上述 Linux 验收，绑定最终候选的实际 SHA。若纳入新的远端提交，重新记录基线与 head，复核变更并重跑受影响验收。文档与原始测试记录各自保留其对应版本，不篡改旧失败、不用新通过覆盖旧记录。
-
-逐节点滚动升级未通过既有二进制校验检查，采用[协调停机方案](../bucket/lifecycle/access-tiering-removal.md#before-upgrading-a-build-with-access-tiering)。实验使用单个 Docker Linux VM 和 tmpfs；正式 tag、包、镜像、跨主机及生产部署仍是独立交付。未验证全部重复 UUID 或运行时统计守恒应如实披露，不能反向引入自动搬回/清理需求或无关重构。
-
-## 9. 执行后记：最终基线、恢复窗口与验收
-
-本次实际交付由 [PR #188](https://github.com/pgsty/silo/pull/188) 承载。选择的远端基线为 `89637554d60c27cfc51d2281d0a4fe15e415f06d`，移植没有包含本地独立 IAM/超时提交 `ebc9937d9`。前三项实现和历史文档逐提交通过 stable patch-id 对照；虚拟补回独立 IAM 差异后，完整树与原审查分支一致。
-
-首次本地 lint 发现新增回调选择分支触发 `gocritic/ifElseChain`，因此追加等价的无表达式 `switch` 改写，保持 marker、指定版本和普通元数据查找的条件顺序及分支体。重新固定的代码候选为 [`41aa84609`](https://github.com/pgsty/silo/commit/41aa84609754769cfb1861d7fd060c2e84182b98)。这一提交上，全量 cmd/internal 得到 6,428 个测试及子测试通过、166 个跳过，50 个有测试的包通过；相关 race 得到 283 个测试及子测试通过。`make build`、全包构建、vet、lint、生成文件及 rebrand/compat 检查均通过。[Go CI](https://github.com/pgsty/silo/actions/runs/34925534139)、[DCO](https://github.com/pgsty/silo/actions/runs/34925534110)、[VulnCheck](https://github.com/pgsty/silo/actions/runs/34925534142) 和[发布流水线的测试运行](https://github.com/pgsty/silo/actions/runs/34925534198)共 11 项检查通过；后者没有发布正式制品。
-
-第一次最终候选停池实验 `dcd5c2e5` 在源版本保全断言后失败：停掉另一池返回 `503 SlowDownRead`，源四盘版本保留；恢复后 DELETE 成功，节点 0/2 的 HEAD/GET 返回 404，节点 1 返回 503。DELETE 成功由脚本已通过的 204 断言确定，原输出没有单独保存该次 DELETE 响应。此次失败如实保留，不能把后来的成功写回原记录。
-
-复核发现，`ListBuckets` 以及位于源池的现存版本 GET，不能证明每个协调节点对另一池的读取连接已经恢复。随后进行了旧基线与候选的同拓扑对照，探测的是从未写入过的随机 UUID，并且在这些探测之前没有执行任何 DELETE：
-
-| 高时间分辨率对照 | 桶列表和现存版本 | 从未写入版本的 HEAD/GET | 观察到的恢复窗口 |
-| --- | --- | --- | --- |
-| `20502a2f`，旧基线 `89637554d` | 三节点均为 200 | 节点 0/2 为 404，节点 1 为 503 SlowDownRead，连续 16 组 | 从重启后的观察循环起算约 1.60–2.50 秒 |
-| `246aaf77`，候选 `41aa84609` | 三节点均为 200 | 同样是节点 1 的 503，连续 11 组 | 约 1.67–2.30 秒 |
-
-两边在缺失版本全节点连续三轮返回 404 后执行 DELETE，均得到 204、全节点 HEAD/GET 404、目标 UUID 在八盘均不存在且其他版本可读。另有两次较低时间分辨率诊断未捕捉到窗口，同样保留；这四次诊断不计入三次升级验收。
-
-这给出了基线在零 DELETE 下的正向复现，证明原恢复条件不足。`getLatestObjectInfoWithIdx` 的读选择函数与基线文本相同：现存副本可以遮蔽另一池的读错误；缺失版本则必须确认所有池，不可读时返回 503。Claude 复核后同意修正实验准备条件并继续验收，明确反对把这一路径的 503 改成 404。最初失败没有瞬时 RPC 全貌，不能逐请求追溯每条连接；这些对照也不能给 `83676ca2` 归因。
-
-修订后的验收在升级/恢复后逐节点探测从未写入的 UUID，记录首次全 404 时刻，要求连续三轮全 404；并列保存各节点 `admin info` 的全盘状态。最多等待 30 秒，超时仍失败，DELETE 之后仍严格要求 204/404，不把 503 纳入通过条件。后续失败保留实验资源供即时取证，再受控清理。
-
-修订准备条件后的独立停池验收 `40a59b3b` 通过：离线 DELETE `503 SlowDownRead`，源四盘版本保留；恢复门约 1.48 秒完成，DELETE `204`，三节点 HEAD/GET `404`，目标在八盘均不存在，另一版本仍可读。恢复早期 `admin info` 中也记录到了各节点不同的离线盘视图，最后恢复为全盘正常。
-
-完整升级验收随后实际进行了三次尝试：
-
-| 尝试 | 运行 | 结果与证据边界 |
-| --- | --- | --- |
-| 1 | `83f88c59` | 准备失败，未启动候选。旧版 rebalance 报 Completed、搬移版本数为 0；源池占用约 6.5%，到平均空闲目标的差值约 3.13%，落入代码既有 5% 容差。增加造数从 64 到 192 个 2 MiB 版本后再试；本次仍计入五次总尝试上限。 |
-| 2，第一轮有效运行 | `ea58d0c5` | 通过。实际 rebalance 中断产生跨八盘的重复版本；停机复制同一数据供旧版对照。旧版 DELETE204 后仍可读；候选 DELETE204 后四节点立即及后续固定间隔 HEAD/GET 均404，八盘目标清除、另一版本四节点可读；旧 ILM/生命周期编辑和真实缓存 v9→v8 通过。 |
-| 3，第二轮有效运行 | `2059bc6b` | **候选失败，阻断合并。** 两阶段逐节点缺失版本连续三轮404、admin info全盘ok之后，DELETE明确204（约12.98ms）；节点2随后的HEAD503/GET503 SlowDownRead，另外三节点404；八盘快照均无目标，首次重试及后续采样全404，其他版本四节点可读。首次失败保留，不因重试恢复改判。 |
-
-第三次尝试发生后停止剩余验收，保留原容器、卷、元数据与响应时序进行诊断。不能把四节点顺序探测中的“节点2异常”直接解释为永久节点故障：它也可能与采样时间有关。随后在保留环境中，对三个额外重复版本做并发、不同顺序的“刚删 UUID / 从未写入 UUID”对照，候选稳定期均为204/404，未复现；旧版克隆数据的双次 DELETE 对照则遇到 `SlowDownWrite`，没有完成其全部断言。这些是诊断结果，不补入有效升级通过计数，也不证明第三次尝试已解释。
-
-为观察逐盘返回，另在独立临时工作树编译仅增加日志的诊断二进制，**没有进入 PR**。它证明了第二层准备检查盲点：`getObjectFileInfo` 的四个响应信号中，可以只有两个实际 `file version not found`，其余两个是被跳过的盘所保留的 `errDiskOngoingReq`；`objectQuorumFromMeta` 的预期读 quorum 为2，因而仍可返回404。同期 `admin info` 汇集各服务器本地盘态为ok，不能证明请求节点到各盘的路径都可用。一次诊断 DELETE 的逐盘返回为 `[nil, nil, drive not found, drive not found]`，达不到写 quorum 3，故返回 `SlowDownWrite`。
-
-Claude 撤回了此前“逐节点三轮404已是最强全池准备条件”的表述，同意这只能证明读 quorum，不能证明全盘可达或写 quorum。诊断给出了候选机制，但**第三次尝试失败瞬间没有逐请求逐盘日志，仍不足以确定其具体原因**；不能把后来稳定期的成功、诊断中的配额不足，或对错误名的解释当成该次故障的直接证据。
-
-## 10. 首次执行的停止位置与恢复资料
-
-首次执行停止时，代码为 `41aa84609`；后续提交只记录执行，不改变已测试的生产代码。当时有效升级运行是一通过、一失败，未满足三次有效运行全部通过的约定；总尝试3次，没有通过继续补跑消耗剩余次数来冲淡失败。`2059bc6b` 和先前的 `83676ca2` 均保持 **OPEN**。Claude 与 Codex 当时的裁定是 **NO-GO for merge**，没有把证据不足升级为“已修复”“既有问题”或“暂态”。后续收尾见第 11 节；实际合并状态以 [#188](https://github.com/pgsty/silo/pull/188) 为准。
-
-主工作区原五文件的完整内容、二进制补丁和 SHA-256 已归档；另有可达 Git 引用 `refs/archive/access-tiering-main-five-files-20260915`，指向快照 `c7fbfc6ada0f0f2abcbe8c0a9681f07e12dafe52`。逐文件核验快照与原已审内容一致。首次停止时因验收未通过，没有执行五文件恢复，也没有移动或改写独立 IAM 提交 `ebc9937d9`。当时唯一待交付候选在 PR 分支，旧变体冻结等待验收裁定。
-
-本机执行资料归档在 `~/.codex/outputs/silo-access-revert-assessment-20260915/final-execution/`：保存了每次尝试、旧/新基线对照、二进制 SHA-256、准备条件、源/目的全盘元数据、诊断补丁、独立评审意见，以及受控清理记录。原始失败记录不覆写；保存的诊断卷内容用于继续调查，不是生产数据或发布制品。
-
-继续推进需要一次能区分机制的取证：在条件可控的升级实验中，于失败请求当时记录每池每盘的真实应答和错误类型，并同时读刚删版本与从未写入版本；明确区分盘面不同步、请求节点的不可用路径与其他原因。若四盘均真实应答仍返回503，应沿错误归约/元数据路径定位；若盘不可用，应查明连接或初始化状态，并验证准备条件。后续成功本身不能关闭本次失败，更不能通过把不可判定的503改成404来满足验收。正式发布、制品和生产部署继续作为独立交付。
-
-## 11. 收尾复核：准备检查、可控机制与独立新轮次
-
-后续收尾没有继续修改生产代码。旧基线 `89637554d` 与候选 `41aa84609` 使用各自的独立诊断构建，仅对测试桶记录逐盘应答；这些日志补丁没有进入 PR。第一轮完整诊断 `53ebe161` 通过但未复现503，仅作为一个样本保留。第二轮 `c951f762` 得到了不同的、可直接解释的失败：两个池的删除调用均记录 `quorum=3 errs=[nil,nil,nil,drive not found]`，DELETE204、所有读样本404，但八盘快照的盘3、7仍保留目标版本。它符合既有写 quorum 契约，并正向证明旧准备门会放行尚未完成挂盘的协调节点；它没有复现或解释 `2059bc6b` 那一次请求。
-
-审查还纠正了两项推断：后续诊断进程在04:19的重连日志不能用于解释03:56的原失败；错误归约按具体错误值计数，不能把“最高同值计数不足”简单等同于“实际应答盘数不足”。原失败之后的全盘快照也不是失败瞬间的原子快照。这些界限继续保留。
-
-准备检查改用服务器已有的 storage trace：从每个 S3 节点，对各自唯一、从未写入的对象执行 `GetObjectTagging`，该路径等待所有盘；按唯一对象名和后端节点、盘路径匹配真实 `storage.ReadVersion` 应答。四节点双池需要每轮32条实际缺失应答，连续三轮完成才满足准备条件；单纯404或 `admin info` 的本地盘状态不够。探针不创建对象，不改变服务器读写语义。trace 输出是多行 JSON 对象流，按流解析；缺失 trace 证据会使检查失败，不能据此假定对应盘健康。
-
-为了比较相同的跨池写入负载，停止真实 rebalance 夹具后复制两份相同数据。旧基线使用 #178 已有的 `If-Match` 条件 DELETE，候选使用普通指定版本 DELETE，二者都处理同一目标的两个池副本。两臂分别完成96次“已删版本／从未写入版本”的 HEAD/GET 对照，均404，目标八盘均清除。准备检查分别耗时约0.60秒和17.08秒。这是一对匹配样本，没有观察到候选专属差异，不是吞吐比较，也不能排除所有可能的故障机制。
-
-`f55967b7` 另用明确制造的重复副本夹具完成因果实验，而非冒充真实 rebalance：pool0 的四盘由一个节点承载并保留 namespace 锁；pool1 的四盘分布到另外四个节点。先停止 pool1 一盘，DELETE204 后直接确认只有该盘残留目标版本。重新接入这份副本，再停止 pool1 两块已删除该版本的盘，保留“一盘返回版本、一盘返回缺失”的读视图。三个被测协调节点的已删版本均返回 `503 SlowDownRead`，从未写入版本均返回404；同一失败请求的逐盘记录明确为 `[drive not found, drive not found, file version not found, nil]`，没有足够的同值应答达到读 quorum。整个实验没有丢失 pool0 的 namespace 锁，也没有将不确定状态改为404。
-
-该因果实验的部分节点重启未在35秒内恢复全部40条访问路径，原实验因此仍记 FAIL；正向机制观察与这个恢复失败分开记录。随后协调重启全部五节点，独立恢复检查通过，五节点 HEAD/GET 均404。这不是滚动恢复已通过的声明，也不追溯原 `2059bc6b` 的逐盘状态。可控复现确定的是故障机制类别，原单次实例继续 **OPEN**。
-
-Claude 复核上述证据后同意显式开启 **R-Upgrade-2**：原轮次的一通过、一失败和总尝试3次保持原样，不合并计数，不静默重置。新轮使用未经诊断修改的 `41aa84609` 二进制，要求三次有效运行全部通过、最多五次总尝试；任何新503或数据不变量失败仍须停止。每次 DELETE 前后均检查32条路径，响应后的即时 HEAD/GET 先于后置准备检查执行，避免等待掩盖短暂错误。
-
-| 新轮次运行 | 完整运行耗时 | 升级后32路径准备耗时 | 验收结果 |
-| --- | --- | --- | --- |
-| `8eb016db` | 90.20秒 | 17.04秒 | PASS |
-| `2a2b7b64` | 80.83秒 | 0.62秒 | PASS |
-| `371e7b9f` | 96.28秒 | 0.60秒 | PASS |
-
-三次均实际走到候选阶段：DELETE204，所有即时及后续 HEAD/GET 样本404，目标在八盘均不存在，其他版本从四节点读回；旧配置、普通生命周期规则编辑及真实缓存 v9→v8 均通过。删除前后各三轮32路径检查也全部通过。准备时间有明显波动，应验证访问路径，不能用固定等待秒数代替检查。这些结果满足修正准备条件后的有界验收；不把有限样本写成“历史503已消失”，不把不同阶段的诊断通过计入新轮次。
-
-新轮仍绑定生产代码 `41aa84609`（Linux 二进制 SHA-256 `28e1339d630a22fa5a0e4659b6182224e81f6cd7cd4079856534390e40a697b1`），后续仅更新文档。合并前须核对源码等价性和最终 CI，按第8节的恢复纪律处理旧五文件，保留独立 IAM 提交。原 `2059bc6b`／`83676ca2`、部分重启恢复边界、单 VM/tmpfs、未验证全部不同重复 UUID 和整份运行时统计守恒继续可见；不为此新增自动搬回、清理服务或读错误降级。
-
-本轮详细资料位于原归档的 `final-execution/closure-20260915/`，包括匹配对照、同请求逐盘日志、`qualification-summary.json`、独立 R-Upgrade-2 账本、诊断补丁和完整卷归档。原失败现场、后续对照及恢复后的卷分别标注时点。临时实验资源在归档验证后清理；这些资料与正式发布制品区分管理。
@@ -1,190 +0,0 @@
-# SILO stack: Go 1.27 compatibility audit
-
-> This is a dated investigation, with the source and runtime boundaries recorded
-> below. It does not establish the current dependency pins or a later release.
-> See [the current changelog](../../CHANGELOG.md) and
-> [component matrix](https://silo.pgsty.com/compatibility/versions/).
-
-
-2026-09-09. Scope: the maintained Server, silo-pkg, mcli, and Console. This extends
-the [OIDC #154 investigation](issue-154.md) to other paths using the same TLS
-configuration and to adjacent standard-library changes. It records local
-validation performed before commit. No pushes, issue comments, releases, or
-production changes were made during the audit.
-
-## Findings and changes
-
-| Component | Finding | Change |
-| --- | --- | --- |
-| Server, baseline `d1105bbb3d4a0afa33b3a4ac11b821235038ed0e` | Eight TLS configuration sites explicitly use the same ML-KEM-containing curve list, overriding `tlsmlkem=0` in Go 1.27. | Remove the eight assignments and obsolete `TLSCurveIDs` helper. Retain Go defaults in HTTP clients, replication, cloud client-certificate transport, both grid links, etcd, and the S3 listener. Add wire-level regression tests. |
-| silo-pkg, baseline `a92c54d` | No built-in explicit PQ curve list. Web environment transport uses defaults; LDAP/OIDC accept caller configuration. | Add a real web-environment TLS regression test and document runtime-default selection. Keep the Go 1.26 library floor. |
-| mcli, baseline `fcd5cad8` | S3/Admin transport and alias/TOFU dialer use default curves. No instance of the Server's override bug found. | Add real S3/Admin transport and alias-dialer handshake tests; update Go/TLS upgrade notes. |
-| Console, baseline `c103d08ec` | IdP, SILO/STS, Prometheus, and webhook clients share a transport using defaults. The HTTPS listener explicitly uses only P-256. | Add real IdP/SILO-client handshake tests; update Go/TLS upgrade notes. Retain the existing listener policy. |
-
-The Server's general external HTTP transport is also used by OpenID discovery
-and JWKS, identity plugins, notification/lambda checks, audit/log webhooks, and
-S3 cloud backends. Fixing only the two OIDC callers would leave those other paths
-affected. The broader correction supersedes the earlier OIDC-only candidate.
-
-The retained defaults follow Go's implementation rather than duplicating its
-`GODEBUG` parser. Go 1.27 explicitly changed the interaction between manually
-selected curves and the `tlsmlkem`/`tlssecpmlkem` default controls. See the
-[Go TLS release notes](https://go.dev/doc/go1.27#crypto/tls).
-
-With no opt-out, default curves additionally include SecP256r1MLKEM768 and
-SecP384r1MLKEM1024. With `GODEBUG=tlsmlkem=0`, hybrid exchanges are disabled for
-default-configured TLS throughout the process. This is an intentional change
-from the old fixed subset. `GODEBUG=tlssecpmlkem=0` disables just the SecP
-hybrids while retaining X25519MLKEM768. TLS versions, cipher-suite policy,
-certificate and hostname checks, client certificates, proxies, and HTTP/2 choices are not
-relaxed by this patch. No automatic fallback after a TLS error is introduced.
-
-## Reproduction and regression evidence
-
-Before changing product code, the new Server test failed for all five tested
-outbound constructors with `tlsmlkem=0`: general external HTTP, internode HTTP,
-replication, cloud client certificates, and etcd. The server observed
-`[X25519MLKEM768 X25519 P256 P384 P521]` in every case. Both TLS 1.2 and TLS 1.3
-peers reproduced the problem. The inbound Server also accepted a PQ-only client
-despite the same opt-out.
-
-After the change, those tests pass on darwin/arm64 and linux/arm64. They exercise
-20 outbound combinations (five constructors, two peer TLS versions, two debug
-settings), plus inbound classical/PQ-only peers with the opt-out enabled and
-disabled. The inbound opt-out case rejects a PQ-only peer while continuing to
-accept P-256. The etcd test exercises its actual TLS configuration, not an etcd
-cluster; the client-certificate test exercises construction/loading, not a full
-mutual-authentication service.
-
-The mc, Console, and silo-pkg handshake tests pass without product code changes.
-All use a trusted synthetic certificate and inspect a real ClientHello. They
-check both disabling and retaining ML-KEM. Console also retains its existing
-unknown-CA, hostname, and endpoint-scoping regression checks.
-
-A freshly source-built complete Linux Server then passed three isolated
-integration scenarios using the existing synthetic IdP fixture:
-
-| Scenario | Result |
-| --- | --- |
-| ML-KEM-intolerant IdP + `tlsmlkem=0` | Discovery/JWKS, IAM, Console, and authenticated Admin calls succeed; login 204 and bucket list 200. |
-| Normal TLS 1.3 IdP, no opt-out | Same successful login, token exchange, STS, session, and bucket-list chain. |
-| Add OIDC to a running Server | Actual Admin API accepts the provider under the compatibility setting. |
-
-Both login scenarios reject modified JWT signatures and a wrong audience: no
-session cookie is issued and authenticated bucket access returns 403. Login
-currently reports these authentication failures as 500; that existing error
-mapping is outside this TLS change. Curl in the same namespace returns HTTP/2
-200. The containers use a pre-existing generic Debian image, `--network none`,
-and no published ports; no Server or Console image was downloaded or used.
-
-The fixture drives real Console HTTP APIs, not rendered browser interaction.
-This is a conditional interoperability reproduction, not proof of the actual
-customer ingress behavior. Grid's existing tests pass, but a production-style
-distributed TLS cluster, external cloud providers, and real etcd/LDAP/Keycloak
-deployments were not exercised. Disabling ML-KEM does not disable new ML-DSA
-signature offers and does not repair an ingress that rejects those offers.
-
-## Other Go changes checked
-
-### macOS root certificates: a confirmed upgrade-visible change
-
-Using the same public synthetic CA and `certs.GetRootCAs`, fresh-process probes
-produce the following results when `SSL_CERT_FILE` points at that CA and
-`SSL_CERT_DIR` points at an empty directory:
-
-| Compiler / main module | CA from environment trusted? | Explicit CA argument trusted? |
-| --- | --- | --- |
-| Go 1.26.5 / `go 1.26.0` | No | Yes |
-| Go 1.27.1 / `go 1.26.0` | No | Yes |
-| Go 1.27.1 / `go 1.27.1` | Yes | Yes |
-
-The Go 1.26 module built by Go 1.27 carries
-`DefaultGODEBUG=...x509sslcertoverrideplatform=0`; explicitly setting that option
-to `1` enables the new behavior. A Go 1.27 application can explicitly set it to
-`0` to recover the previous platform behavior. The diagnostic source is
-[cert-roots.go](issue-154/cert-roots.go). The consuming application's defaults
-apply to library calls as well, so silo-pkg's older `go` directive does not
-prevent the behavior in Server, mc, or Console.
-
-This is expected standard-library behavior, not a reason to silently discard
-configured CA variables or skip verification. Setting either variable replaces
-Keychain trust with on-disk roots and Go's verifier; stale or incomplete paths
-can break previously trusted connections. Unset inherited variables to restore
-Keychain trust; explicit additional CAs still work. The three application
-READMEs and the package README now document this. The package's Windows loader enumerates
-the native ROOT store directly and does not call `SystemCertPool`; it does not
-inherit this particular new setting. Windows behavior was reviewed in source,
-not runtime-tested.
-
-### JSON, HTTP, timers, and other compatibility controls
-
- **JSON:** checked owned JSON error handling and exercised policy/condition,
-  config, and authentication tests. `quick` uses typed `SyntaxError` and
-  `UnmarshalTypeError`; no owned decision depending on changed standard JSON
-  error text was found. silo-pkg's full suite passes with both Go compilers;
-  mc's command and Console's API/auth suites pass under Go 1.27. No broad
-  `nojsonv2` opt-out or serialization rewrite is justified by these results.
- **HTTP response closing:** Go 1.27's standard-library drain is bounded at
-  256 KiB and 50 ms. mc's actual early-close/cancel regression passes, including
-  the compressed S3 Select stream. Existing owned drain helpers can still have
-  independent timeout concerns; those are not newly caused by this Go change.
- **ALPN and custom connections:** mc's TLS dialer returns a real `*tls.Conn`;
-  its deadline wrapper is on the TCP dial path. No new accidental HTTP/2 opt-in
-  from the expanded `ConnectionState` interface support was identified in
-  these transports. Console keeps its existing transport policy.
- **Removed switches:** no maintained runtime/config reliance on `tlsrsakex`,
-  `tls3des`, `tls10server`, `x509keypairleaf`, or `asynctimerchan` was found.
-  Existing explicit cipher policies continue to be explicit. Timer uses in
-  package certificate reload and license refresh do not depend on buffered
-  timer-channel length/capacity. Unix EOF error changes do not expose a matching
-  owned error-type assumption on the reviewed Unix-socket paths.
- **Platform floor:** application READMEs now identify macOS 13 as the minimum
-  for Go 1.27 binaries. No Windows, old macOS, or PowerPC runtime claim is made.
-
-## Validation and delivery
-
- Server: focused TLS tests on macOS and Linux; macOS race run; crypto, HTTP,
-  grid, and all `internal/config/...` tests; full Linux Server build and the
-  three integration scenarios above.
- silo-pkg: `make test` (lint and full race suite) on Go 1.27.1; full suite on
-  Go 1.26.5, preserving the library floor.
- mc: complete `./cmd` suite, including the new TLS test and existing S3 Select
-  early-cancel test.
- Console: complete `./api/... ./pkg/...` suites, including identity and STS
-  validation and the new handshake tests.
-
-Go lint reports zero issues in all four repositories. Server's optional spelling
-check is skipped because `typos` is not installed. Its lint target used the
-already-installed matching golangci-lint v2.13.1 after a redundant download was
-stopped; mc's lint was rerun serially after the global linter lock prevented the
-first concurrent attempt. These tooling retries did not require source changes.
-
-An independent Fable 5.1 Max review reproduced the old-code failures and ran the
-new TLS tests with the race detector. Its shuffled mc and Console suites passed.
-One shuffled silo-pkg run failed the untouched `certs.TestValidPairAfterWrite`;
-that test passed three plain reruns, the same shuffle seed, and the full certs
-package rerun. The certs package runs in a separate test binary from the changed
-env tests; this was classified as an existing timing flake.
-
-The review also found that committing the diagnostic fixture made rebrand CI
-count synthetic IdP paths as product routes. The guard now excludes
-`docs/investigations/`; the product compatibility baseline is unchanged.
-The full proposed file set passes the guard, and a negative control adding a
-product route still fails. After the review corrections, a fresh Linux Server
-build passed all three integration scenarios again; the new binary identity and
-rerun results are recorded in the evidence file's `post_review_validation`.
-
-All source and dependency choices remain those of the maintained PGSTY stack.
-`go.mod` and `go.sum` are unchanged in every repository. The Server uses its
-existing pinned Console/pkg/mc modules; the companion changes are tests and
-documentation, so no replacement graph or unpublished dependency version is
-needed to build the runtime fix.
-
-Validation used separate worktrees for Server, silo-pkg, mc, and Console, leaving
-the original repositories and their `main` branches untouched. Build identities,
-fixture results, and command outcomes are recorded in
-[go127-stack-evidence.json](go127-stack-evidence.json). Absolute paths and branch
-names in the captured evidence identify the environment at recording time.
-The recorded module graph identifies the dependencies selected for those
-builds; `go.sum` also retains checksums for unselected versions. Local image IDs
-and temporary paths are historical evidence, not portable setup instructions.
@@ -1,534 +0,0 @@
-{
-  "date": "2026-09-11",
-  "runs": [
-    {
-      "runid": "silo-v1-0806-9c53f14c",
-      "version": "0806",
-      "image": "sha256:29a498b24669cae1fed11c1a2fb2b3d73c68829a0a9c0b14e71b386671d38fac",
-      "nodes": 4,
-      "drives": 4,
-      "filesystem": "Linux tmpfs named volumes, held mounted across server restarts",
-      "drive_bytes": 268435456,
-      "status": "PASS",
-      "phases": [
-        {
-          "phase": "startup-admin",
-          "seconds_from_start": 4.157,
-          "first_online_by_coordinator": [
-            4.037,
-            4.078,
-            4.118,
-            3.737
-          ]
-        },
-        {
-          "phase": "startup-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.239,
-          "seconds_from_start": 4.396,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.096,
-            0.116,
-            0.127,
-            0.138
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.162,
-            0.194,
-            0.217,
-            0.239
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "full-restart-admin",
-          "seconds_from_start": 2.084,
-          "first_online_by_coordinator": [
-            1.937,
-            1.976,
-            2.016,
-            1.643
-          ]
-        },
-        {
-          "phase": "full-restart-canary",
-          "attempts": 45,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 14.449,
-          "seconds_from_start": 16.534,
-          "first_put_seconds_after_admin_by_coordinator": [
-            14.315,
-            13.573,
-            13.953,
-            0.05
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            14.373,
-            14.396,
-            14.426,
-            14.449
-          ],
-          "transient_error_count": 129,
-          "first_transient_errors": [
-            {
-              "attempt": 1,
-              "operation": "put",
-              "node": 0,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 1,
-              "operation": "put",
-              "node": 1,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 1,
-              "operation": "put",
-              "node": 2,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 2,
-              "operation": "put",
-              "node": 0,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            }
-          ]
-        },
-        {
-          "phase": "readback-15s",
-          "seconds_after_canary": 17.596,
-          "objects": 55,
-          "reads": 220,
-          "errors": []
-        },
-        {
-          "phase": "readback-30s",
-          "seconds_after_canary": 31.357,
-          "objects": 55,
-          "reads": 220,
-          "errors": []
-        },
-        {
-          "phase": "readback-60s",
-          "seconds_after_canary": 61.566,
-          "objects": 55,
-          "reads": 220,
-          "errors": []
-        },
-        {
-          "phase": "one-node-outage",
-          "existing_read": true,
-          "put": true,
-          "readers": 3
-        },
-        {
-          "phase": "rejoin-admin",
-          "seconds_from_start": 1.382,
-          "first_online_by_coordinator": [
-            1.27,
-            1.307,
-            1.343,
-            1.382
-          ]
-        },
-        {
-          "phase": "rejoin-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.158,
-          "seconds_from_start": 1.541,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.013,
-            0.029,
-            0.039,
-            0.05
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.077,
-            0.102,
-            0.132,
-            0.158
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "final-readback",
-          "seconds_after_canary": 1.71,
-          "objects": 60,
-          "reads": 240,
-          "errors": []
-        }
-      ],
-      "acknowledged_objects": 60,
-      "version_ids_recorded": true,
-      "sha256_recorded": true
-    },
-    {
-      "runid": "silo-v1-0903-a457573c",
-      "version": "0903",
-      "image": "sha256:b616a0cf8cb281e7e6bb3c9b1fb53875b4016a2878223925541c18f82d6c5ca3",
-      "nodes": 4,
-      "drives": 4,
-      "filesystem": "Linux tmpfs named volumes, held mounted across server restarts",
-      "drive_bytes": 268435456,
-      "status": "PASS",
-      "phases": [
-        {
-          "phase": "startup-admin",
-          "seconds_from_start": 4.271,
-          "first_online_by_coordinator": [
-            4.158,
-            4.194,
-            3.827,
-            3.867
-          ]
-        },
-        {
-          "phase": "startup-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.2,
-          "seconds_from_start": 4.471,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.054,
-            0.068,
-            0.081,
-            0.092
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.118,
-            0.146,
-            0.167,
-            0.2
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "full-restart-admin",
-          "seconds_from_start": 2.284,
-          "first_online_by_coordinator": [
-            2.143,
-            2.193,
-            0.849,
-            0.902
-          ]
-        },
-        {
-          "phase": "full-restart-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.191,
-          "seconds_from_start": 2.476,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.016,
-            0.029,
-            0.044,
-            0.057
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.093,
-            0.124,
-            0.158,
-            0.192
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "readback-15s",
-          "seconds_after_canary": 15.235,
-          "objects": 8,
-          "reads": 32,
-          "errors": []
-        },
-        {
-          "phase": "readback-30s",
-          "seconds_after_canary": 30.356,
-          "objects": 8,
-          "reads": 32,
-          "errors": []
-        },
-        {
-          "phase": "readback-60s",
-          "seconds_after_canary": 60.192,
-          "objects": 8,
-          "reads": 32,
-          "errors": []
-        },
-        {
-          "phase": "one-node-outage",
-          "existing_read": true,
-          "put": true,
-          "readers": 3
-        },
-        {
-          "phase": "rejoin-admin",
-          "seconds_from_start": 0.903,
-          "first_online_by_coordinator": [
-            0.791,
-            0.829,
-            0.865,
-            0.903
-          ]
-        },
-        {
-          "phase": "rejoin-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.141,
-          "seconds_from_start": 1.044,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.012,
-            0.022,
-            0.035,
-            0.045
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.069,
-            0.096,
-            0.12,
-            0.141
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "final-readback",
-          "seconds_after_canary": 0.301,
-          "objects": 13,
-          "reads": 52,
-          "errors": []
-        }
-      ],
-      "acknowledged_objects": 13,
-      "version_ids_recorded": true,
-      "sha256_recorded": true
-    },
-    {
-      "runid": "silo-v1-current-6a14eb18",
-      "version": "current",
-      "image": "pgsty/d12a:build",
-      "image_id": "sha256:307af7711e2e04ab75759cb42a1eef45c43c4404894c0e30dd19f742b107b922",
-      "platform": "linux/arm64",
-      "nodes": 4,
-      "drives": 4,
-      "filesystem": "Linux tmpfs named volumes, held mounted across server restarts",
-      "drive_bytes": 268435456,
-      "status": "PASS",
-      "binary_sha256": "1e4cd7b78ecfa1b0cf28f1961d48a220a712c6be1fd3a01b60ee99aec62f04a4",
-      "source_revision": "b32f2d9dd01a383a9991d34ffe248063463a031d+pool-consistency",
-      "phases": [
-        {
-          "phase": "startup-admin",
-          "seconds_from_start": 7.374,
-          "first_online_by_coordinator": [
-            7.207,
-            6.677,
-            6.74,
-            6.8
-          ]
-        },
-        {
-          "phase": "startup-canary",
-          "attempts": 1,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 0.479,
-          "seconds_from_start": 7.854,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.266,
-            0.279,
-            0.292,
-            0.305
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            0.336,
-            0.378,
-            0.431,
-            0.479
-          ],
-          "transient_error_count": 0,
-          "first_transient_errors": []
-        },
-        {
-          "phase": "full-restart-admin",
-          "seconds_from_start": 2.461,
-          "first_online_by_coordinator": [
-            2.264,
-            1.414,
-            2.396,
-            2.461
-          ]
-        },
-        {
-          "phase": "full-restart-canary",
-          "attempts": 36,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 14.489,
-          "seconds_from_start": 16.951,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.013,
-            0.025,
-            14.301,
-            0.046
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            14.358,
-            14.394,
-            14.435,
-            14.489
-          ],
-          "transient_error_count": 104,
-          "first_transient_errors": [
-            {
-              "attempt": 1,
-              "operation": "put",
-              "node": 2,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 1,
-              "operation": "get",
-              "node": 2,
-              "key": "full-restart-attempt-1-node-0",
-              "error": "An error occurred (SlowDownRead) when calling the GetObject operation (reached max retries: 0): Resource requested is unreadable, please reduce your request rate"
-            },
-            {
-              "attempt": 1,
-              "operation": "get",
-              "node": 2,
-              "key": "full-restart-attempt-1-node-3",
-              "error": "An error occurred (SlowDownRead) when calling the GetObject operation (reached max retries: 0): Resource requested is unreadable, please reduce your request rate"
-            },
-            {
-              "attempt": 2,
-              "operation": "put",
-              "node": 2,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            }
-          ]
-        },
-        {
-          "phase": "readback-15s",
-          "started_seconds_after_canary": 15.0,
-          "seconds_after_canary": 22.006,
-          "objects": 113,
-          "reads": 452,
-          "errors": []
-        },
-        {
-          "phase": "readback-30s",
-          "started_seconds_after_canary": 30.003,
-          "seconds_after_canary": 36.571,
-          "objects": 113,
-          "reads": 452,
-          "errors": []
-        },
-        {
-          "phase": "readback-60s",
-          "started_seconds_after_canary": 60.0,
-          "seconds_after_canary": 65.743,
-          "objects": 113,
-          "reads": 452,
-          "errors": []
-        },
-        {
-          "phase": "one-node-outage",
-          "existing_read": true,
-          "put": true,
-          "readers": 3
-        },
-        {
-          "phase": "rejoin-admin",
-          "seconds_from_start": 2.49,
-          "first_online_by_coordinator": [
-            1.165,
-            2.417,
-            1.646,
-            2.087
-          ]
-        },
-        {
-          "phase": "rejoin-canary",
-          "attempts": 37,
-          "puts": 4,
-          "gets": 16,
-          "hard_deadline_seconds": 60,
-          "gate_seconds": 13.977,
-          "seconds_from_start": 16.468,
-          "first_put_seconds_after_admin_by_coordinator": [
-            0.013,
-            0.024,
-            0.035,
-            13.889
-          ],
-          "first_four_reads_seconds_after_admin_by_coordinator": [
-            13.909,
-            13.927,
-            13.953,
-            13.977
-          ],
-          "transient_error_count": 36,
-          "first_transient_errors": [
-            {
-              "attempt": 1,
-              "operation": "put",
-              "node": 3,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 2,
-              "operation": "put",
-              "node": 3,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 3,
-              "operation": "put",
-              "node": 3,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            },
-            {
-              "attempt": 4,
-              "operation": "put",
-              "node": 3,
-              "error": "An error occurred (SlowDownWrite) when calling the PutObject operation (reached max retries: 0): Resource requested is unwritable, please reduce your request rate"
-            }
-          ]
-        },
-        {
-          "phase": "final-readback",
-          "started_seconds_after_canary": 0.0,
-          "seconds_after_canary": 6.507,
-          "objects": 226,
-          "reads": 904,
-          "errors": []
-        }
-      ],
-      "acknowledged_objects": 226,
-      "version_ids_recorded": true,
-      "sha256_recorded": true
-    }
-  ]
-}
@@ -1,327 +0,0 @@
-#!/usr/bin/env python3
-"""Bounded four-container restart/readback acceptance for pgsty/silo#116."""
-import argparse
-import hashlib
-import json
-import os
-from pathlib import Path
-import secrets
-import shutil
-import signal
-import socket
-import subprocess
-import time
-import uuid
-from concurrent.futures import ThreadPoolExecutor
-
-import boto3
-from botocore.config import Config
-
-ROOT = None
-MCLI = None
-CURRENT_BINARY = None
-SOURCE_REVISION = None
-IMAGES = {
-    '0806': 'pgsty/silo:RELEASE.2026-08-06T00-00-00Z',
-    '0903': 'pgsty/silo:RELEASE.2026-09-03T13-18-01Z',
-    'current': 'pgsty/d12a:build',
-}
-
-
-def docker(*args, timeout=40, check=True):
-    return subprocess.run(['docker', *args], check=check, capture_output=True, text=True, timeout=timeout)
-
-
-class Deadline(BaseException):
-    pass
-
-
-def bounded(seconds, operation):
-    def expired(*_):
-        raise Deadline(f'hard deadline of {seconds}s exceeded')
-    old = signal.signal(signal.SIGALRM, expired)
-    signal.setitimer(signal.ITIMER_REAL, seconds)
-    try:
-        return operation(time.monotonic() + seconds)
-    finally:
-        signal.setitimer(signal.ITIMER_REAL, 0)
-        signal.signal(signal.SIGALRM, old)
-
-
-def remaining(deadline):
-    value = deadline - time.monotonic()
-    if value <= 0:
-        raise Deadline('absolute deadline exceeded')
-    return value
-
-
-def run(version):
-    image_info = json.loads(docker('image', 'inspect', IMAGES[version]).stdout)[0]
-    runid = f'silo-v1-{version}-{uuid.uuid4().hex[:8]}'
-    out = ROOT / runid
-    out.mkdir(mode=0o700)
-    user, password = 'local116', secrets.token_urlsafe(24)
-    envfile = out / 'credentials.env'
-    envfile.write_text(f'MINIO_ROOT_USER={user}\nMINIO_ROOT_PASSWORD={password}\nMINIO_CI_CD=1\nMINIO_BROWSER=off\nGOMAXPROCS=2\n')
-    envfile.chmod(0o600)
-    env = {k: v for k, v in os.environ.items() if not k.startswith(('MINIO_', 'SILO_', 'MC_'))
-           and k.lower() not in {'http_proxy', 'https_proxy', 'all_proxy', 'no_proxy'}}
-    nodes = [f'{runid}-n{i}' for i in range(4)]
-    sockets = [socket.socket() for _ in nodes]
-    for sock in sockets:
-        sock.bind(('127.0.0.1', 0))
-    ports = [sock.getsockname()[1] for sock in sockets]
-    for sock in sockets:
-        sock.close()
-    volumes = [n + '-data' for n in nodes]
-    endpoints, ledger = [], []
-    result = {'runid': runid, 'version': version, 'image': IMAGES[version],
-              'image_id': image_info['Id'], 'platform': image_info['Os']+'/'+image_info['Architecture'],
-              'nodes': 4,
-              'drives': 4, 'filesystem': 'Linux tmpfs named volumes, held mounted across server restarts',
-              'drive_bytes': 268435456, 'phases': [], 'status': 'RUNNING'}
-    if version == 'current':
-        result['binary_sha256'] = hashlib.sha256(CURRENT_BINARY.read_bytes()).hexdigest()
-        result['source_revision'] = SOURCE_REVISION
-    bucket = 'canary-' + uuid.uuid4().hex[:10]
-
-    def save(event=None):
-        if event is not None:
-            result['phases'].append(event)
-            compact = {k: (len(v) if k in ('transient_errors', 'errors') else v) for k, v in event.items()}
-            print(json.dumps({'version': version, **compact}), flush=True)
-        (out / 'result.json').write_text(json.dumps(result, indent=2) + '\n')
-        (out / 'acknowledged.json').write_text(json.dumps(ledger, indent=2) + '\n')
-
-    def parallel(fn, values):
-        with ThreadPoolExecutor(max_workers=4) as pool:
-            return list(pool.map(fn, values))
-
-    def client(i, deadline):
-        timeout = remaining(deadline)
-        return boto3.client('s3', endpoint_url=endpoints[i], aws_access_key_id=user,
-                            aws_secret_access_key=password, region_name='us-east-1',
-                            config=Config(proxies={}, signature_version='s3v4', s3={'addressing_style': 'path'},
-                                          retries={'total_max_attempts': 1}, connect_timeout=timeout,
-                                          read_timeout=timeout, request_checksum_calculation='when_required',
-                                          response_checksum_validation='when_required'))
-
-    def admin_gate(origin, phase):
-        def check(deadline):
-            first = [None] * 4
-            while True:
-                states = []
-                for i in range(4):
-                    try:
-                        p = subprocess.run([str(MCLI), '--config-dir', str(out / 'mcli'), '--json',
-                                            'admin', 'info', f'n{i}'], env=env, capture_output=True,
-                                           text=True, timeout=min(5, remaining(deadline)))
-                        info = json.loads(p.stdout)['info']
-                        servers = info['servers']
-                        ok = len(servers) == 4 and all(s.get('state') == 'online' and s.get('drives')
-                             and all(d.get('state') == 'ok' for d in s['drives'])
-                             and all(v == 'online' for v in s.get('network', {}).values()) for s in servers)
-                        if ok:
-                            (out / f'{phase}-admin-{i}.json').write_text(json.dumps(info, indent=2) + '\n')
-                            if first[i] is None:
-                                first[i] = round(time.monotonic() - origin, 3)
-                        states.append(ok)
-                    except (Exception,):
-                        states.append(False)
-                if all(states):
-                    event = {'phase': phase + '-admin', 'seconds_from_start': round(time.monotonic()-origin, 3),
-                             'first_online_by_coordinator': first}
-                    save(event)
-                    return time.monotonic()
-                time.sleep(min(.25, remaining(deadline)))
-        return bounded(90, check)
-
-    def canary(phase, origin, admin_time, setup=False):
-        def check(deadline):
-            started = time.monotonic()
-            first_put, first_reads = [None] * 4, [None] * 4
-            errors, attempt, setup_done = [], 0, not setup
-            result['active_canary'] = {'phase': phase, 'errors': errors}
-            while True:
-                attempt += 1
-                remaining(deadline)
-                if not setup_done:
-                    try:
-                        try:
-                            client(0, deadline).create_bucket(Bucket=bucket)
-                        except Exception as e:
-                            if 'BucketAlreadyOwnedByYou' not in str(e):
-                                raise
-                        client(0, deadline).put_bucket_versioning(Bucket=bucket, VersioningConfiguration={'Status': 'Enabled'})
-                        setup_done = True
-                    except Exception as e:
-                        errors.append({'attempt': attempt, 'operation': 'setup', 'error': str(e)[:250]})
-                        time.sleep(min(.25, remaining(deadline)))
-                        continue
-                acked = []
-                for i in range(4):
-                    key = f'{phase}-attempt-{attempt}-node-{i}'
-                    payload = (key + '\n').encode() * 16384
-                    try:
-                        vid = client(i, deadline).put_object(Bucket=bucket, Key=key, Body=payload)['VersionId']
-                        entry = {'phase': phase, 'key': key, 'version': vid, 'bytes': len(payload),
-                                 'sha256': hashlib.sha256(payload).hexdigest(), 'writer': i,
-                                 'ack_seconds_from_start': round(time.monotonic()-origin, 3)}
-                        ledger.append(entry)
-                        save()  # Persist every acknowledged write, including failed rounds.
-                        acked.append(entry)
-                        if first_put[i] is None:
-                            first_put[i] = round(time.monotonic()-admin_time, 3)
-                    except Exception as e:
-                        errors.append({'attempt': attempt, 'operation': 'put', 'node': i, 'error': str(e)[:250]})
-                reads = 0
-                for i in range(4):
-                    own_reads = 0
-                    for entry in acked:
-                        try:
-                            verify(i, entry, deadline)
-                            reads += 1
-                            own_reads += 1
-                        except Exception as e:
-                            errors.append({'attempt': attempt, 'operation': 'get', 'node': i,
-                                           'key': entry['key'], 'error': str(e)[:250]})
-                    if own_reads == 4 and first_reads[i] is None:
-                        first_reads[i] = round(time.monotonic()-admin_time, 3)
-                remaining(deadline)
-                if len(acked) == 4 and reads == 16:
-                    result.pop('active_canary', None)
-                    save({'phase': phase + '-canary', 'attempts': attempt, 'puts': 4, 'gets': 16,
-                          'hard_deadline_seconds': 60, 'gate_seconds': round(time.monotonic()-started, 3),
-                          'seconds_from_start': round(time.monotonic()-origin, 3),
-                          'first_put_seconds_after_admin_by_coordinator': first_put,
-                          'first_four_reads_seconds_after_admin_by_coordinator': first_reads, 'transient_errors': errors})
-                    return time.monotonic()
-                (out / f'{phase}-canary-errors.json').write_text(json.dumps(errors, indent=2) + '\n')
-                if attempt == 1:
-                    print(json.dumps({'version': version, 'phase': phase, 'first_attempt_errors': errors}), flush=True)
-                time.sleep(min(.25, remaining(deadline)))
-        return bounded(60, check)
-
-    def verify(i, entry, deadline):
-        got = client(i, deadline).get_object(Bucket=bucket, Key=entry['key'], VersionId=entry['version'])
-        try:
-            data = got['Body'].read()
-        finally:
-            got['Body'].close()
-        assert len(data) == entry['bytes'] and hashlib.sha256(data).hexdigest() == entry['sha256'], entry['key']
-        assert got.get('VersionId') == entry['version'], entry['key']
-        remaining(deadline)
-
-    def readback(label, origin):
-        def check(deadline):
-            started = time.monotonic()
-            errors = []
-            for entry in ledger:
-                for i in range(4):
-                    try:
-                        verify(i, entry, deadline)
-                    except Exception as e:
-                        errors.append({'key': entry['key'], 'node': i, 'error': str(e)[:250]})
-            save({'phase': label, 'started_seconds_after_canary': round(started-origin, 3),
-                  'seconds_after_canary': round(time.monotonic()-origin, 3),
-                  'objects': len(ledger), 'reads': len(ledger)*4, 'errors': errors})
-            return errors
-        return bounded(60, check)
-
-    try:
-        docker('network', 'create', runid)
-        for volume in volumes:
-            docker('volume', 'create', '--driver', 'local', '--opt', 'type=tmpfs',
-                   '--opt', 'device=tmpfs', '--opt', 'o=size=256m', volume)
-        mounts = [arg for i, v in enumerate(volumes) for arg in ('--mount', f'type=volume,source={v},target=/keep/{i}')]
-        docker('run', '-d', '--pull=never', '--network', 'none', '--name', runid + '-keeper',
-               *mounts, '--entrypoint', 'sleep', 'alpine:3.23', '1800')
-        urls = [f'http://{n}:9000/data' for n in nodes]
-        def create(i):
-            extra = []
-            if version == 'current':
-                extra = ['--mount', f'type=bind,source={CURRENT_BINARY},target=/lab/silo,readonly',
-                         '--entrypoint', '/lab/silo']
-            docker('create', '--pull=never', '--name', nodes[i], '--network', runid,
-                   '--hostname', nodes[i], '--cpus', '2', '--memory', '3g', '--env-file', str(envfile),
-                   '--mount', f'type=volume,source={volumes[i]},target=/data',
-                   '-p', f'127.0.0.1:{ports[i]}:9000', *extra, image_info['Id'], 'server', '--address', ':9000',
-                   '--console-address', ':9001', *urls)
-        parallel(create, range(4))
-        start = time.monotonic()
-        parallel(lambda n: docker('start', n), nodes)
-        for i, n in enumerate(nodes):
-            endpoint = 'http://' + docker('port', n, '9000/tcp').stdout.strip()
-            endpoints.append(endpoint)
-            env[f'MC_HOST_n{i}'] = endpoint.replace('http://', f'http://{user}:{password}@')
-        admin = admin_gate(start, 'startup')
-        canary('startup', start, admin, setup=True)
-        parallel(lambda n: docker('stop', '-t', '10', n), nodes)
-        start = time.monotonic()
-        parallel(lambda n: docker('start', n), nodes)
-        admin = admin_gate(start, 'full-restart')
-        gate = canary('full-restart', start, admin)
-        read_errors = []
-        for after in (15, 30, 60):
-            time.sleep(max(0, gate + after - time.monotonic()))
-            read_errors.extend(readback(f'readback-{after}s', gate))
-        assert not read_errors, 'acknowledged object readback failure; see result.json'
-        docker('stop', '-t', '10', nodes[3])
-        def outage(deadline):
-            verify(0, ledger[0], deadline)
-            payload = b'acknowledged with one Linux node offline' * 16384
-            key = 'one-node-outage'
-            vid = client(0, deadline).put_object(Bucket=bucket, Key=key, Body=payload)['VersionId']
-            entry = {'phase': 'outage', 'key': key, 'version': vid, 'bytes': len(payload),
-                     'sha256': hashlib.sha256(payload).hexdigest(), 'writer': 0}
-            ledger.append(entry)
-            save()
-            for i in range(3):
-                verify(i, entry, deadline)
-            save({'phase': 'one-node-outage', 'existing_read': True, 'put': True, 'readers': 3})
-        bounded(60, outage)
-        start = time.monotonic()
-        docker('start', nodes[3])
-        admin = admin_gate(start, 'rejoin')
-        gate = canary('rejoin', start, admin)
-        assert not readback('final-readback', gate)
-        result['status'] = 'PASS'
-    except BaseException as e:
-        result['status'] = 'FAIL'
-        result['error'] = f'{type(e).__name__}: {e}'
-        raise
-    finally:
-        save()
-        for n in nodes:
-            log = docker('logs', n, check=False)
-            (out / (n + '.log')).write_text(log.stdout + log.stderr)
-            docker('rm', '-f', n, check=False)
-        docker('rm', '-f', runid + '-keeper', check=False)
-        for v in volumes:
-            docker('volume', 'rm', v, check=False)
-        docker('network', 'rm', runid, check=False)
-        envfile.unlink(missing_ok=True)
-        print(json.dumps({'version': version, 'status': result['status'], 'evidence': str(out)}), flush=True)
-
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument('versions', nargs='+', choices=list(IMAGES))
-    parser.add_argument('--output', type=Path, required=True, help='directory for retained evidence')
-    parser.add_argument('--mcli', default=shutil.which('mcli'), help='native mcli executable')
-    parser.add_argument('--current-binary', type=Path, help='Linux binary matching the Docker architecture')
-    parser.add_argument('--source-revision', help='Git revision of --current-binary')
-    parser.add_argument('--current-image', default=IMAGES['current'], help='cached Linux base image for current binary')
-    args = parser.parse_args()
-    if not args.mcli or not Path(args.mcli).is_file():
-        parser.error('--mcli must point to an executable file')
-    if 'current' in args.versions and (not args.current_binary or not args.current_binary.is_file()):
-        parser.error('current requires --current-binary')
-    ROOT = args.output.resolve()
-    ROOT.mkdir(mode=0o700, parents=True, exist_ok=True)
-    MCLI = Path(args.mcli).resolve()
-    CURRENT_BINARY = args.current_binary.resolve() if args.current_binary else None
-    SOURCE_REVISION = args.source_revision
-    IMAGES['current'] = args.current_image
-    for version in args.versions:
-        run(version)
@@ -1,409 +0,0 @@
-# SILO #154：OIDC discovery 连接重置调查
-
-> This is a dated investigation, with the source and runtime boundaries recorded
-> below. It does not establish the current dependency pins or a later release.
-> See [the current changelog](../../CHANGELOG.md) and
-> [component matrix](https://silo.pgsty.com/compatibility/versions/).
-
-
-前两轮调查时间：2026-09-09；公开 issue 最后核对于 07:53 UTC。第二轮补充同源码、同依赖、不同 Go 工具链的 Linux 完整 Server 对照和候选补丁认证链路验证。
-
-**后续更新：用户已授权扩展至整个 SILO 技术栈并修复。现已确认 Server 其他 TLS 路径也存在同类覆盖问题，并在产品工作区完成统一使用 Go 默认曲线的修复。当前实现、验证和交付状态见 [全栈调查](go127-stack.md)。下文保留前两轮的诊断与当时的 OIDC 局部候选；“未修改产品”和“不要扩大范围”等表述仅适用于当时的调查阶段，局部候选已被后续全路径修复取代。**
-
-## 判断与处理顺序
-
-**目前可以确认是 Server 发出的 discovery GET 失败，随后 IAM 初始化等待，Console 初始化也被延后；还不能确认真实连接由谁、在哪个协议阶段重置。优先调查 SILO/Go 客户端与 IdP 前置 TLS 终止器、代理或 WAF 的互操作。** 现有证据不支持将其定性为证书错误、Keycloak 配置错误、JWT 校验错误或 `coreos/go-oidc` 回归。
-
-有两项与版本相关、可在本地验证的 TLS 差异：
-
-1. **Go 1.27 改变了 `GODEBUG=tlsmlkem=0` 与显式 `CurvePreferences` 的关系。** SILO 两个版本都显式包含 X25519MLKEM768。旧版 Go 1.26.5 会根据该环境选项移除它；Go 1.27.1 保留显式配置。在模拟拒绝 ML-KEM 的入口上，能重现“相同选项下旧 Server 启动成功，新 Server 持续 reset”。**客户是否设置过这个选项尚未知，不能把条件性复现当成客户根因。**
-2. **Go 1.27 的 ClientHello 新增 ML-DSA 签名算法。** 没有上述环境选项时，新旧版本也会发送不同的握手。人为拒绝新算法编号的入口同样能产生旧成功、新失败。真实入口是否存在这种行为尚未知。
-
-另外，HTTP User-Agent 从 `MinIO` 变成了 `Silo`；若连接在 TLS 完成、GET 发出之后被重置，应优先查 WAF、User-Agent 规则和 HTTP 路由，而不是继续调整 TLS。
-
-最短路径：**先在故障进程所在环境拿到实际 Go 版本、是否设置 `tlsmlkem=0`、目标 IP 和 TLS 完成与否；再针对已证实的分支处理。** 优先修正入口兼容性或错误路由。若确认是上述环境选项失效，只为 OpenID 出站请求恢复 Go 默认曲线选择，是目前最小的代码候选。现阶段不宜做全局 TLS 改动或整体依赖回退。
-
-## 第二轮结论：已隔离 Go 因素，局部候选修复通过 Linux 验证
-
-**可以复现一种由 Go 1.26.5 → 1.27.1 单独触发的兼容性回归。建议保留 Go、针对已确认分支修复；整体回退只用于临时恢复服务。** 这里的“已确认”指本地实验机制，仍不等于已经确认客户入口的根因。
-
-四个完整 Server 都从本地源码编译为 `linux/arm64`，在现成通用 Debian 12 基础镜像的 `--network none` 容器中运行；Server、合成 IdP、Console 和 curl 共用同一 loopback 网络，没有暴露端口。没有下载或运行 Server/Console 镜像。旧源码为 `d88f46ccee345a9c2fabe2d221d9a9e56bc11aec`，当前源码为 `d1105bbb3d4a0afa33b3a4ac11b821235038ed0e`。
-
-旧源码两次构建使用完全相同的 `go.mod`、`go.sum` 和实际链接模块版本，仅替换编译器；当前源码与候选补丁构建的依赖图也完全相同。实际二进制 SHA-256、`go version -m` 依赖和构建选项保存在 [Linux 证据](issue-154/linux-evidence.json)。
-
-下表的入口**人为设置为见到 X25519MLKEM768 就发 TCP RST**，Server 都设置 `GODEBUG=tlsmlkem=0`：
-
-| 源码与编译器 | Server 初始 ClientHello | 完整 Server 结果 | 同容器 curl |
-| --- | --- | --- | --- |
-| 同一份旧源码 + Go 1.26.5 | 275 字节，无 ML-KEM | discovery、JWKS、IAM、Console 正常；cluster 200 | HTTP/2 200 |
-| 同一份旧源码 + Go 1.27.1 | 1509 字节，仍包含 ML-KEM | `connection reset by peer`；IAM 等待；cluster 503；Console 未启动 | HTTP/2 200 |
-| 当前源码 + Go 1.27.1 | 1509 字节，仍包含 ML-KEM | 同样失败 | HTTP/2 200 |
-| 当前源码 + 局部候选补丁 + Go 1.27.1 | 287 字节，无 ML-KEM | 完整启动及合成 OIDC 登录成功 | HTTP/2 200 |
-
-这将该条件下的回归定位到工具链行为，而非 Console、pkg、mc 或 `coreos/go-oidc` 升级。Go 1.27 的发布说明明确将此作为有意改变：`tlsmlkem` / `tlssecpmlkem` 只控制默认曲线集合，显式指定的集合可以继续启用这些算法。SILO 现有曲线列表显式包含该算法，因而原来的兼容开关在这条路径失效。[Go 1.27 crypto/tls 说明](https://go.dev/doc/go1.27#crypto/tls)。
-
-共完成 **12 个 Linux 场景**，其中失败用例按预期失败：
-
- 正常 TLS 1.2 / P-256 / RSA / AES-256-GCM 的 IdP，旧源码用两个 Go 版本编译均正常，证明不是 Go 1.27 普遍无法连接这套 TLS。
- 上表四个对照均符合预期；候选补丁如果不设置 `tlsmlkem=0`，仍被 ML-KEM 拒绝规则拦截。补丁恢复显式兼容选项的作用，不会自行关闭后量子算法。
- 候选补丁在上述 TLS 1.2 兼容场景、以及无 GODEBUG 的正常 TLS 1.3 场景，都完成 Console 登录信息获取 → IdP authorization redirect → callback → token exchange → STS 凭据 → Console 会话 → 桶列表读取。登录 API 为 204，桶列表为 200。
- 两个登录场景分别提交错误签名与错误 audience 的 JWT，登录返回 500、桶列表返回 403，没有生成会话 cookie。这里只确认认证未放行，没有将现有 500 状态码行为改为另一项修复。
- 不信任 CA 时，候选 Server 仍因 `x509: certificate signed by unknown authority` 停在 IAM 初始化。没有关闭证书校验。
- 不配置 OIDC 启动后，经实际 Admin API 添加同一合成 provider：当前源码失败且返回 reset；候选补丁成功。
- 入口改为拒绝 ML-DSA 签名编号，候选补丁加 `tlsmlkem=0` 仍失败。这是另一条机制，当前补丁没有解决它。
-
-认证流程由 Python 驱动真实 Console HTTP API；IdP 使用一次性合成授权码和自行签发的实验 JWT，没有客户账户。没有执行浏览器页面交互、登出、真实 Keycloak 或生产数据升级测试。TLS 1.3 对照实际协商 TLS 1.3/P-256，不是所有新增后量子曲线的互操作覆盖。
-
-### 修复与回退的取舍
-
-[候选补丁](issue-154/openid-default-curves.patch) 只有三个文件：增加 OpenID 专用 transport helper，将其 `TLSClientConfig.CurvePreferences` 设为 `nil`，再替换 IAM 初始化和 OpenID 配置校验两个调用点。测试时仅应用于隔离源码副本，当前产品工作区未应用；Go 版本与所有依赖保持不变。它让 Go 默认策略和现有兼容开关接管外部 IdP 的密钥交换，其他 TLS 参数继续来自现有构造函数。
-
-如果客户证据确认此分支，建议采用该局部修复，并在客户实际入口复验。如果客户没有设置 `tlsmlkem=0`，它不能单独解释旧版成功：旧版默认也发送 ML-KEM。此时应继续区分 ML-DSA、其他握手变化、HTTP/WAF 规则与网络路径；不把此补丁直接宣称为 #154 的完整修复。
-
-**不建议将当前产品直接改回 Go 1.26。** 实测以 Go 1.26.7 和 `GOTOOLCHAIN=local` 读取当前源码即被 `go.mod requires go >= 1.27.1` 拒绝；所固定的 Server、Console、mc 均声明 Go 1.27.1。回退需要进一步调整这些模块及可能的传递依赖，不是只换一个编译器版本。这个报错证明当前依赖图不能原样回编，并不证明经过额外适配后绝对无法回编。报告中已恢复服务的旧版可作为临时运行状态，不能把这种处置等同于完成兼容修复。
-
-第二轮的运行脚本是 [run-linux.py](issue-154/run-linux.py)，结果是 [linux-evidence.json](issue-154/linux-evidence.json)，具体构建和运行步骤见文末。没有 issue 评论、提交、推送、合并或发布。
-
-候选副本另通过 Go 1.27.1、darwin/arm64、CGO 关闭的 `go test -mod=readonly -count=1 ./internal/http ./internal/config/identity/openid`；补丁可以干净应用到调查基线。这些包级测试与上述 Linux 集成验证分别记录，不将其当成 Linux 单元测试结果。
-
-## 范围、版本与公开证据
-
- 初始工作区干净，处于 detached HEAD；调查分支为 `codex/investigate-oidc-154`，基线与远端 main 均为 `d1105bbb3d4a0afa33b3a4ac11b821235038ed0e`。
- 已读取 `/Users/vonng/pgsty/silo/AGENTS.md` 及工作区适用说明。维护范围为 PGSTY 的 Server、Console、mc、silo-pkg；上游 MinIO 仅作参考。
- 第一轮 Server 与 transport 探针为本地源码构建的 darwin/arm64 程序，第二轮完整 Server 交叉构建为 linux/arm64；均使用 `CGO_ENABLED=0 GOWORK=off`。fixture 与 Admin 辅助程序也由本机 Go 构建。历史源码使用 `git archive` 导入隔离临时目录。没有下载或运行 Server/Console Docker 镜像，没有访问客户端点，没有改动真实服务或数据，没有评论 issue、推送、合并或发布。
- 产品源码、`go.mod`、`go.sum` 未修改。本目录中的 Go 文件是显式运行的调查工具，带 `//go:build ignore`，不进入正常构建。
-
-| 项目 | 旧版：2026-08-04 | 报告故障版：2026-09-03 | 调查时 main |
-| --- | --- | --- | --- |
-| 完整 tag | `RELEASE.2026-08-04T00-00-00Z` | `RELEASE.2026-09-03T13-18-01Z` | 无新 release 声明 |
-| 源码 commit | `d88f46ccee345a9c2fabe2d221d9a9e56bc11aec` | `9b11dc9469e650815b775cb47b039610644f5da4` | `d1105bbb3d4a0afa33b3a4ac11b821235038ed0e` |
-| go.mod / Docker 构建定义 | Go 1.26.5 | Go 1.27.1 | Go 1.27.1 |
-| 本地 Server / 探针实际编译器 | Go 1.26.5 | Go 1.27.1 | Go 1.27.1 |
-| Console replacement | `v0.0.0-20260804042150-b952a1202869` | `v0.0.0-20260903111932-464a59d73ada` | `v0.0.0-20260908142700-c103d08ec36a` |
-| mc replacement | `v0.0.0-20260801042411-ad10a2a10b76` | `v0.0.0-20260903063637-a2ef95c035d9` | `v0.0.0-20260909015522-fcd5cad8247f` |
-| PGSTY silo-pkg | v3.11.0，替换历史 minio/pkg 路径 | v3.13.2，直接依赖 | v3.13.3，直接依赖 |
-| coreos/go-oidc/v3 | v3.17.0 | v3.21.0 | v3.21.0 |
-| x/crypto | v0.54.0 | v0.56.0 | v0.56.0 |
-| x/net | v0.57.0 | v0.58.0 | v0.58.0 |
-| x/oauth2 | v0.36.0 | v0.36.0 | v0.36.0 |
-
-版本依据：[旧版 go.mod](https://github.com/pgsty/silo/blob/d88f46ccee345a9c2fabe2d221d9a9e56bc11aec/go.mod)、[故障版 go.mod](https://github.com/pgsty/silo/blob/9b11dc9469e650815b775cb47b039610644f5da4/go.mod)、[本次 main go.mod](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/go.mod)。还核对了各 tag 的 `Dockerfile.goreleaser` 和 release workflow。它们声明相应 Go 构建版本、禁用 CGO；历史归档、本次探针和 Server 均用 `go version -m` 核实实际编译器与 replacement。**这些不是客户实际镜像 digest 或其内二进制的取证，后者仍需 `--version` / build info 确认。**
-
-[Issue #154](https://github.com/pgsty/silo/issues/154) 当前 OPEN，最后更新时间 `2026-09-08T05:30:42Z`，评论数为 0。报告包含升级后 Server 初始化失败、旧版回退恢复、测试实例添加 OIDC 失败，以及 curl 成功的输出。
-
-curl 那次连接验证了所收到的证书链，并选择 TLS 1.2、`ECDHE-RSA-AES256-GCM-SHA384`、P-256 和 HTTP/2；解析得到两个 IPv4，记录中实际访问了其中一个。**公开命令使用 `docker run --rm` 新建容器，并非 `docker exec` 进入原故障容器；同镜像不能证明同网络命名空间、环境变量、CA 挂载、DNS 结果或出口。** 域名、realm、IP 已脱敏，本次不推测其真实值。
-
-## 实际请求链
-
-### IAM 与配置校验
-
-```text
-Server startup
-  IAMSys.Init
-    openid.LookupConfig
-      parseDiscoveryDoc: GET .well-known/openid-configuration
-      PopulatePublicKey: GET discovery 中的 jwks_uri
-    IAM store 初始化
-    Console 初始化
-
-Console 添加 OIDC
-  AdminClient.AddOrUpdateIDPConfig
-    Server addOrUpdateIDPHandler
-      validateConfig(identity_openid)
-        同一个 openid.LookupConfig / NewHTTPTransport
-```
-
-`parseDiscoveryDoc` 是 SILO 自己的实现，用标准库 `http.Client` 发 GET，收到成功响应后才解码 JSON；这次日志中的 `Get ... read tcp ... reset` 发生在该调用返回响应之前，不能进一步区分 TLS 与 HTTP。请求本身不需要客户 client secret、token 或私钥。[IAM 调用点](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/iam.go#L278-L288)、[discovery 实现](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/internal/config/identity/openid/jwt.go#L261-L285)、[JWKS 实现](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/internal/config/identity/openid/jwt.go#L89-L110)。
-
-Console 的添加表单通过 Admin API 触发 Server 配置校验。校验成功才写入配置；本地已重现该 API 返回同样的 reset，恢复 IdP 后再次创建成功并要求重启。[Console 调用](https://github.com/pgsty/silo-console/blob/c103d08ec36a/api/admin_idp.go#L79-L114)、[Admin 客户端](https://github.com/pgsty/silo-console/blob/c103d08ec36a/api/client-admin.go#L593-L595)、[Server 校验和保存边界](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/admin-handlers-idp-config.go#L127-L150)、[OpenID 配置校验](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/config-current.go#L353-L359)。
-
-`coreos/go-oidc.NewProvider` 在 Server 中的使用是 `MockOpenIDTestUserInteraction` 测试辅助函数，不是这次 IAM discovery 调用链。升级此依赖不能单独解释或修复该 GET。`crypto/tls` 和这里的 `net/http` 属于 Go 标准库，不能用 go.mod 中 `x/crypto`、`x/net` 的版本代替它们的实际行为。[辅助函数](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/utils.go#L994-L1012)。
-
-### transport 参数与环境
-
-| 项目 | 实际行为及意义 |
-| --- | --- |
-| 构造 | `NewHTTPTransport()` → `NewHTTPTransportWithTimeout(time.Minute)` → `xhttp.ConnSettings`。每次 IAM 初始化重试都会重新构造。 |
-| TLS 版本 | 未显式设置 Min/Max；所比较 Go 工具链的正常默认范围是 TLS 1.2–1.3。证书、主机名验证开启。 |
-| 密码套件 | 显式 `TLSCiphersBackwardCompatible()`，包含 curl 成功使用的 ECDHE-RSA/AES-256-GCM。没有理由为本 issue 添加旧 RSA、3DES 或 SHA-1 例外。 |
-| 曲线 | 显式 `{X25519MLKEM768, P256, X25519, P384, P521}`，两个 release 和当前 main 相同。实际上线顺序还受 Go 实现控制，见实验。 |
-| ALPN / HTTP | `EnableHTTP2=false`，同时设置自定义 TLS config 和 DialContext；实测 ClientHello 没有 ALPN，GET 使用 HTTP/1.1。三个版本一致。`GODEBUG=http2client=0` 对此基线路径无修复价值。 |
-| 代理 | `http.ProxyFromEnvironment`；HTTPS URL 使用 `HTTPS_PROXY` / `https_proxy` 与 `NO_PROXY` / `no_proxy`。这两版标准库均优先非空大写值，不使用 `ALL_PROXY`；设置在进程内缓存。不能根据 curl 的路由推断它。注意 go.mod 的 x/net v0.58.0 代码不是这里所用的标准库 vendored 实现。 |
-| DNS | `globalDNSCache.LookupHost`，dnscache v0.1.1；默认刷新窗口在容器/Kubernetes 为 30 秒，其他环境为 10 分钟，可配置。按返回地址逐个尝试 TCP，首次 TCP 成功就返回；随后 TLS/HTTP 失败不会回到这个循环尝试另一 IP。代码注释称随机选择，但该循环没有 shuffle。 |
-| Linux TCP 参数 | 使用 SILO 的自定义 dialer，包含 TCP fast open/keepalive 等设置，部分参数来自 Server CLI，包括 interface、buffer、user timeout。第二轮执行了完整 Linux Server 的正常 CLI 初始化，但只验证 loopback；不能排除实际出口设备、路由或非默认 CLI 参数的交互。 |
-| CA | `silo-pkg/certs.GetRootCAs` 加载系统根、Kubernetes CA 目录和 `certs/CAs`；Server 还加入自己的公开服务证书。相关 pkg CA 加载实现未在这次版本比较中变化；实际文件和路径仍可能因部署变化不同。 |
-| 超时 | TCP 拨号 5 秒，TLS 握手 10 秒，响应头 1 分钟；discovery/JWKS 的 Client 没有总超时，discovery 使用的 Request 也没有调用方 context。响应体停滞不受响应头超时保护。 |
-| 复用 | keep-alive 开启，idle 15 秒，TLS session cache 100。一次初始化内 discovery 与 JWKS 可复用连接；初始化重试的新 transport 没有旧连接或 session。持续首次握手失败不能用清理空闲连接解释。 |
-| 请求标识 | UA 包含产品、OS、架构、模式和构建信息，产品名从 `MinIO` 变为 `Silo`；传输层禁用自动压缩。UA 规则只能在 HTTPS 被终止、HTTP 请求可见之后起作用。 |
-
-源码：[构造与超时](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/utils.go#L651-L670)、[HTTP/TLS 参数](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/internal/http/transports.go#L44-L94)、[密码套件与曲线](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/internal/crypto/crypto.go#L52-L78)、[DNS 拨号循环](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/internal/http/dial_dnscache.go#L42-L84)、[缓存刷新](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/common-main.go#L551-L578)、[CA 加载](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/server-main.go#L383-L395)、[UA](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/update.go#L228-L266)。
-
-比较旧 tag 与故障 tag，OpenID discovery、transport、曲线实现的变动仅为 pkg 导入路径调整；IAM 另有品牌日志变更。比较故障 tag 与本次 main，这些关键实现没有变动。因此不能把当前依赖推进当成 #154 已解决的证据。
-
-**Console 登录阶段是另一条出站路径。** 当前及故障版 Console 通过 `GetConsoleHTTPClient` / `GlobalTransport` 再获取 discovery、交换令牌，TLS config 未指定 CurvePreferences，仍做完整证书验证；不能把“添加配置走 Server”推广为所有 Console OIDC 请求都走 Server transport。任何修复最终都必须验证完整登录。`CONSOLE_MINIO_SERVER_TLS_SKIP_VERIFY` 只针对 SILO 端点，不能作为 IdP 修复。[Console transport](https://github.com/pgsty/silo-console/blob/c103d08ec36a/api/config.go#L68-L121)、[IdP 客户端](https://github.com/pgsty/silo-console/blob/c103d08ec36a/api/tls.go#L59-L70)、[Console discovery](https://github.com/pgsty/silo-console/blob/c103d08ec36a/pkg/auth/idp/oauth2/provider.go#L428-L449)。
-
-## 本地实验与能排除的假设
-
-工具与原始结果放在 [issue-154/](issue-154/)：
-
- `probe.go` 直接调用 `cmd.NewHTTPTransport()`，用同版本 `certs.GetRootCAs` 装入实验 CA；记录 TCP、TLS、HTTP 阶段。诊断参数仅修改被比较的一项。
- `fixture.go` 仅监听 IPv4 loopback；使用即时生成的 RSA 测试证书与专用 CA，提供 discovery 和有效 JWKS。正常基线限制 TLS 1.2、P-256、`TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384`，支持 HTTP/1.1 和 HTTP/2。
- `admin-check.go` 仅允许 loopback Server，验证 Console 使用的 Admin API；使用本地虚构配置和临时凭据。
- `evidence.json` 保存版本、握手参数、分阶段 trace 和完整 Server 实验结果；无客户信息、私钥或 token。
- 第二轮扩展 fixture 的合成授权码、token 和 JWKS 流程；`run-linux.py` 驱动 Linux 完整 Server 及 Console API，`linux-evidence.json` 保存十二组场景和构建身份。敏感运行值不写入结果。
-
-### 正常服务及握手差异
-
-三个源码版本的实际 transport 都能通过完整证书验证，以 TLS 1.2 / AES-256-GCM / HTTP/1.1 获取测试文档。**Go 1.27 本身并非不能连接 TLS 1.2、RSA 证书或这套密码套件。** 正常服务也接受诊断性的 HTTP/2。
-
-| 构建与选项 | fixture 读到的初始握手字节数 | 支持的 group ID | 新增 ML-DSA 签名编号 |
-| --- | ---: | --- | --- |
-| 旧源码 + Go 1.26.5，默认 | 1497 | 4588, 29, 23, 24, 25 | 无 |
-| 故障源码 + Go 1.27.1，默认 | 1509 | 同上 | 0x0904, 0x0905, 0x0906 |
-| 当前源码 + Go 1.27.1，默认 | 1509 | 同上 | 同上 |
-| 旧源码 + Go 1.26.5，`tlsmlkem=0` | 275 | 29, 23, 24, 25 | 无 |
-| 故障/当前源码 + Go 1.27.1，`tlsmlkem=0` | 1509 | 4588, 29, 23, 24, 25 | 有 |
-| 旧源码、旧依赖，只改为 Go 1.27.1 | 1509 | 4588, 29, 23, 24, 25 | 有；`tlsmlkem=0` 也不再移除 4588 |
-
-4588 是 X25519MLKEM768，29 是 X25519，23/24/25 是 P-256/384/521。这些字节数包含本地 fixture 收到的 TLS record，测试 URL 是 IP，没有 DNS SNI；不能直接当成客户网络中的包长或 MTU 证据。默认新旧差异约 12 字节，没有证据支持“本次才突然出现巨大 ML-KEM 握手”的说法。设置上述环境选项时的差异则显著不同。
-
-旧源码保留旧依赖、仅更换 Go 编译器后，行为随编译器变化，隔离了本次发现与 silo-pkg/Console 版本更新之间的关系。Go 1.27 官方说明也明确记录显式曲线配置不再受这些默认值开关限制，并新增 ML-DSA 支持。[Go 1.27 发布说明](https://go.dev/doc/go1.27)。本地进一步核对了两版 `crypto/tls/defaults.go`、`common.go` 的 `curvePreferences`/`supportsCurve` 和 `handshake_client.go`。
-
-### 主动拒绝与对照结果
-
-下列拒绝规则是人为设置的模型，**只证明机制可以产生相同症状，不证明真实 Keycloak 或其入口使用这些规则**。
-
-| fixture 规则 | 对照结果 | 能支持的结论 |
-| --- | --- | --- |
-| ClientHello 带 ML-KEM 就发送 TCP RST | 旧版默认也失败；旧版 + `tlsmlkem=0` 成功；故障版 + 同选项失败；故障版显式 classical 曲线成功 | 需要旧环境选项或其他变化，才能用该机制解释升级回归。仅说 ML-KEM 不兼容不充分。 |
-| ClientHello 带 ML-DSA 编号就 RST | 旧版默认成功；故障版默认/仅 classical 都失败；故障版 TLS 1.2-only 成功 | 新的签名算法列表是另一种可区分机制。ML-DSA 与 ML-KEM 不是同一项。TLS 1.2-only 会同时改变多项 ClientHello，成功不等于唯一定位 ML-DSA。 |
-| 必须提供 h2 ALPN | 旧、新 Server transport 默认都失败；`-h2` 成功 | 可以解释 curl 与 Server 的不同，单独不能解释新旧版本差异。 |
-| TLS 完成后，对 `Silo` UA 的 GET 发送 RST | 同一个新 transport，`MinIO` UA 成功、`Silo` UA 失败 | 报告的外层 `Get ... reset` 错误也可能来自 HTTP 层，必须先判断 TLS 是否完成。 |
-| 不信任测试 CA | 返回 `*tls.CertificateVerificationError` / x509 类错误 | 与人为 RST 的错误不同；没有证据要求跳过证书验证。仍须比较客户实际连接收到的链。 |
-| 正常连接复用 | 第二个请求 `reused=true`；关闭 idle 连接后重新建 TCP 可恢复 TLS session | keep-alive 与 TLS session 复用是两件事。首次新 transport 就失败时，此分支优先级低。 |
-
-### 完整 Server、IAM 和恢复
-
-使用三个本地编译的完整 Server，独立空数据/配置目录、独立 CA 和临时凭据；未登录真实 IdP。
-
-| 场景 | 实际结果 |
-| --- | --- |
-| 旧版、故障版连接正常 IdP | discovery + JWKS 成功；一条 TLS 连接；cluster health 200；Console HTTP 200；Admin ListUsers 成功。 |
-| 当前版连接持续 reset 的 IdP | IAM 持续等待；cluster health 503，`X-Minio-Server-Status: iam-offline`；Console 端口尚未提供服务；5 秒限时的 Admin ListUsers 未完成。 |
-| 上述场景恢复 IdP，不重启 Server | 约 0.43 秒后 cluster/Console 200，ListUsers 成功。该时间是单次本地样本，不是恢复 SLA。 |
-| 旧版 + `tlsmlkem=0`，IdP 拒绝 ML-KEM | 完整启动成功。 |
-| 故障版 + `tlsmlkem=0`，相同拒绝规则 | IAM 等待、cluster 503；撤掉规则后约 1.38 秒内完整恢复，无需重启。 |
-| 当前版 discovery 成功、JWKS 返回 503 | 同样阻塞 IAM；JWKS 恢复后约 0.33 秒内恢复。只验证 discovery 200 不够。 |
-| 当前版先不配置 OIDC，再经 Admin API 添加 | IdP reset 时返回同型 `Get ... read tcp ... connection reset by peer`；恢复后同名创建成功，`restart=true`。失败校验没有保存该 provider。 |
-
-在本地 IAM 阻塞的场景中，`/minio/health/live` 和 `/minio/health/ready` **仍为 200**。判断这次恢复应使用 `/minio/health/cluster` 并验证受认证操作和 Console，不能只看 ready。源码上 cluster 的 `checkHealth` 检查 IAM，而 ready 没有此检查；这是已有行为，本文不扩展为一次健康检查重构。[health 检查](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/healthcheck-handler.go#L32-L65)、[ready 检查](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/healthcheck-handler.go#L129-L186)。
-
-现有 IAM 重试间隔随机为 0–3 秒，GET 本身还会占用网络等待时间；初始化成功前不会继续创建 IAM store，Console 又在 IAM 调用返回后启动。外部连接恢复可由现有重试自行恢复；这不意味着靠增加重试能修复持续不兼容。缺少整体请求期限、请求取消的部分是另外一个可单独修复的启动健壮性问题。[重试逻辑](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/iam.go#L342-L365)、[Console 启动顺序](https://github.com/pgsty/silo/blob/d1105bbb3d4a0afa33b3a4ac11b821235038ed0e/cmd/server-main.go#L1007-L1027)。
-
-已通过：`CGO_ENABLED=0 GOWORK=off go test -mod=readonly -count=1 ./internal/http ./internal/config/identity/openid`。这些测试没有替代真实 Keycloak 登录或 Linux 故障网络的验证。
-
-## 最少补充证据与诊断命令
-
-先收集第一轮，按结果才展开后续。所有请求只取公开 discovery，不需要客户端密钥、密码、token 或私钥。不要求环境变量全量导出、配置导出或证书私钥。
-
-### 第一轮：运行身份、网络一致性、协议阶段
-
-在**已有故障容器/Pod 的实际网络环境**中运行；旧版也做同样检查。不要以新建默认网络容器替代。如果实际进程经启动脚本修改过环境，探针也应使用修改后的相关环境和同一 CA 挂载。
-
-```sh
-# 选择实际 Server 可执行文件；记录输出中的版本、Go、OS/架构。
-silo --version
-# 旧镜像的程序名可能是 minio。
-
-# Linux：PID 设为实际 Server 进程 PID，不预设一定是 1。
-# 只输出 Go 调试选项以及相关环境项是否存在，不输出代理凭据/地址。
-tr '\0' '\n' < "/proc/$PID/environ" | awk '
-  /^GODEBUG=/ { print; next }
-  /^(HTTP_PROXY|HTTPS_PROXY|NO_PROXY|ALL_PROXY|http_proxy|https_proxy|no_proxy|all_proxy|SSL_CERT_FILE|SSL_CERT_DIR)=/ {
-    split($0, a, "="); print a[1] "=<set>"
-  }'
-
-# OIDC_URL 仅在本地设为原 config_url；URL 不应带凭据或令牌。
-export OIDC_URL
-curl --http1.1 --connect-timeout 5 --max-time 20 -sS -o /dev/null \
-  -w 'ip=%{remote_ip} http=%{http_version} status=%{http_code} verify=%{ssl_verify_result}\n' "$OIDC_URL"
-curl --http2 --connect-timeout 5 --max-time 20 -sS -o /dev/null \
-  -w 'ip=%{remote_ip} http=%{http_version} status=%{http_code} verify=%{ssl_verify_result}\n' "$OIDC_URL"
-
-# 由维护者从对应源码构建的探针；OIDC_CA 是与 Server 相同的 CAs 目录。
-./oidc-probe -ca "$OIDC_CA"
-```
-
-探针从 `OIDC_URL` 读取 URL。URL、响应体、请求头不会打印；输出的 IP 可按一致映射替换为 IP-A/IP-B。它使用真实 Server transport 构造函数，但添加 20 秒总期限、不跟随重定向、限制响应体读取为 1 MiB，并使用 `issue-154-probe` UA；因此是定位连接阶段的工具，不是完整 OIDC 流程。它没有执行 Server 的 CLI 初始化，也不继承该进程的 `--interface`、socket buffer、TCP user timeout 或存量 DNS 缓存；若使用这些非默认参数，必须对齐后才能归因。若 Server 将自己的公开服务证书也作为根信任，需要把相同公开证书加入探针的临时 CA 目录。
-
-读结果的方法：
-
- `tls_start` 后 `tls_done ... err=reset`：先查 TLS ClientHello、代理 CONNECT 或 TLS 终止设备；尚不能从客户端确定 RST 由终端还是中间设备发出。
- `tls_done ... err=none`、`wrote_request` 后 reset：检查 HTTP/WAF/UA/入口路由。此时更换证书信任或密钥交换没有针对性。
- 出现 x509 类错误：比较该进程实际收到的证书链、SNI、系统 CA 和自定义 CA；保持验证开启。
- 两次 curl 的 IP 不同，或探针目标与 curl 不同：先做同 IP 对照。curl HTTP/1.1 成功也不代表 Go ClientHello 相同。
- 仅 curl h2 成功：确认实际协商的 `http_version` 是 2，再检查入口 HTTP/1.1 支持；不要先全局启用 Server HTTP/2。
-
-### 只对命中的分支做 A/B
-
-```sh
-# TLS 阶段失败：仅移除 hybrid key exchange，仍支持 TLS 1.3 和验证证书。
-./oidc-probe -ca "$OIDC_CA" -classical
-
-# 如果旧进程确实使用 tlsmlkem=0，验证最小候选是否恢复其效果。
-# 应保留原 GODEBUG 的其他相关选项；以下假定没有需要保留的其他值。
-GODEBUG=tlsmlkem=0 ./oidc-probe -ca "$OIDC_CA" -default-curves
-
-# 只有 classical 仍失败时，作为鉴别实验测试 TLS 1.2-only。
-./oidc-probe -ca "$OIDC_CA" -tls12
-
-# 只有 HTTP/ALPN 对照指向此分支时才测试。
-./oidc-probe -ca "$OIDC_CA" -h2
-
-# HTTP 阶段才失败：UA_OLD / UA_NEW 是两版实际请求的完整 UA，非密钥。
-./oidc-probe -ca "$OIDC_CA" -ua "$UA_OLD"
-./oidc-probe -ca "$OIDC_CA" -ua "$UA_NEW"
-```
-
-若 first-hop 使用代理，先比较两进程的代理选择和 `NO_PROXY`，不公开含密码的代理 URL。只在明确允许直连的部署中使用 `-direct`。确认直连后，可针对每个 DNS 地址运行以下两项，保留原 URL 主机名与 SNI，不能直接把 HTTPS URL 改成 IP：
-
-```sh
-curl --noproxy '*' --resolve "$OIDC_HOST:443:$IP_A" --http1.1 \
-  --connect-timeout 5 --max-time 20 -sS -o /dev/null \
-  -w 'ip=%{remote_ip} status=%{http_code} verify=%{ssl_verify_result}\n' "$OIDC_URL"
-./oidc-probe -ca "$OIDC_CA" -direct -ip "$IP_A"
-# 对 IP-B 重复；端口不是 443 时据实修改 --resolve。
-```
-
-仅第二次请求失败时才补 `-n 2` 与 `-n 2 -fresh`；后者重建 TCP，但同一 transport 的 TLS session cache 仍保留。若仍无法区分，下一步要的是入口侧同一时间窗口的握手失败原因/命中规则，或由客户自行脱敏后的 ClientHello 参数和 RST 阶段，不是完整认证流量包。
-
-## 条件性最小修复方案
-
-### 1. 路由、代理或入口策略差异已证实
-
-优先统一有问题的 IdP 入口、修正具体域名的代理/NO_PROXY 或后端节点配置；更新错误拒绝合法 ClientHello 的 TLS 终止器/WAF。证书仍按原 hostname 和有效 CA 验证，OIDC issuer 不随意改名。若是 `Silo` UA 被规则拒绝，调整该规则；不把全产品 UA 改回 MinIO。
-
-这是配置层处理，可以不改 SILO。工作量取决于入口归属；成功标准是原版故障二进制在原网络环境中完成 discovery、JWKS 和完整登录，而非仅 curl 200。
-
-### 2. 证实旧环境依赖 `tlsmlkem=0`，且 classical / default-curves 对照成功
-
-最小代码候选是为外部 OpenID 请求使用 Go 的默认曲线集合，恢复已有 GODEBUG 选项的效果；保留当前 trust pool、SNI、密码套件、超时、代理及 HTTP/1.1 行为。候选代码如下，**已在隔离副本完成上述验证，尚未应用到产品工作区**：
-
-```go
-func NewOpenIDHTTPTransport() *http.Transport {
-    tr := NewHTTPTransport()
-    tr.TLSClientConfig.CurvePreferences = nil
-    return tr
-}
-```
-
-当时的候选只将 `cmd/iam.go` 和 `cmd/config-current.go` 两处 OpenID 调用改用该 helper，通过现有 UA wrapper 传入 `openid.LookupConfig`。后续排查确认节点互联、复制、远端存储等连接也存在同样的问题，此局部方案已被 [SILO 跨组件修复](go127-stack.md) 取代；请勿再应用下面归档的 OIDC-only 补丁。
-
-候选已在真实 transport 探针及第二轮完整 Linux Server 中验证：`CurvePreferences=nil` + Go 1.27.1 + `tlsmlkem=0` 会移除 ML-KEM，并通过 `reject-mlkem` fixture、配置添加和合成登录。它仍不能通过 `reject-mldsa` fixture，说明此方案针对的是一个明确分支。
-
-影响需要明确：
-
- 不设置 GODEBUG 时会采用完整 Go 默认集合，实测额外提供 SecP256r1MLKEM768 和 SecP384r1MLKEM1024；这也是一项兼容性变化。第二轮默认 TLS 1.3 登录通过，仍不能称为完全无行为变化或所有曲线均已覆盖。
- `GODEBUG` 是进程级设置，其他使用 Go 默认曲线的客户端也可能受到影响；本次 helper 的改动范围是 OpenID，但该环境选项本身不是逐 provider 开关。
- 禁用 hybrid 后仍有标准 ECDHE/TLS 和证书验证，失去的是相应后量子密钥交换保护。优先修正入口，临时兼容设置应有撤销条件；不能自动遇到 reset 就降级重试。
- 当前 Console IdP transport 本来使用默认曲线，同一进程中的该 GODEBUG 策略可以与之保持一致；第二轮合成登录链路已经验证，真实 IdP 和浏览器验收仍待进行。
- 隔离源码副本中的候选实现及上述本地验证已完成；真实端点 A/B 证据、生产适用性和发布是尚未完成的步骤。
-
-### 3. 未设置上述选项，或 classical 仍失败
-
-不要套用方案 2。若同 IP、同代理、同 UA 下仅 Go 1.27 失败，优先收集入口对新签名算法/扩展的处理证据。`-tls12` 成功只能缩小到 ClientHello/TLS 特征集合；它同时移除 hybrid key share 和 TLS 1.3 特有签名编号，不能唯一证明 ML-DSA。
-
-优先更新入口或获得 Go 上游可复现用例。只有真实证据证明 TLS 1.2 兼容模式是必要且有效、入口又短期无法处理时，才评估**明确限定于该 IdP**的临时 TLS 1.2 选项，并保留现代 ECDHE/AEAD 和证书验证。不要把 `MaxVersion=TLS12` 全局写死，不引入自定义 ClientHello/TLS 栈或未受支持的“关闭 ML-DSA”环境选项。该分支尚不足以确定补丁或工期。
-
-### 4. 启动健壮性可单独改进
-
-如需小幅改善可诊断性，应单独给 discovery/JWKS 加明确阶段标识和有限总请求期限，让 IAM 重试可感知取消；在请求完成前发生 stall 时及时释放资源。日志避免输出 client secret/token，错误保留原始 cause。保留已配置 OIDC 的失败关闭行为及 IdP 恢复后的自动初始化，不自动关闭 OIDC，也不把无限重试包装成根因修复。
-
-这类改动约 0.5–1 个工程日，可分别回归 stalled body、取消、重试后恢复；它不会使持续 RST 的 TLS 连接成功，不应替代前面鉴别。
-
-## 临时处置与验收边界
-
-当前可沿用报告中已经恢复服务的旧版回退状态，尽快完成上述定位。不要将旧版本长期保留视为解决方案，也不要为了它整体降级新版本依赖。本次空数据实验不证明任意生产数据/配置的降级兼容；再次切换版本前应沿既有备份和升级边界操作。
-
-修复应按以下范围验证，均从 PGSTY 源码本地构建：
-
-1. 对确认的分支增加最小回归：实际 OpenID transport、匹配的 ClientHello/入口拒绝条件、默认设置与明确 opt-out；确保通用和 internode transport 不变。
-2. TLS 1.2（报告密码套件）与 TLS 1.3、有效自定义 CA、错误 CA 和 hostname 拒绝；适用时覆盖代理及多地址入口。不放宽证书、JWT 签名或 audience 等认证校验。
-3. 完整 Server 的 discovery 与 JWKS、IdP 中断后恢复、cluster health、受认证 Admin 操作；Console 添加配置、浏览器重定向、回调、令牌交换、STS 授权和登出。mc 添加/读取配置路径应一致。
-4. 在实际 Linux 容器/Pod 网络和每个 IdP 入口地址复验。证据应绑定最终 Server/Console/pkg/mc commit 和编译器；发布、镜像及生产可用性属于后续独立验收。
-
-第二轮完成 Linux loopback 上的合成 OAuth token/STS 登录；本次没有真实 Keycloak、真实 Linux 故障路径、浏览器页面或生产数据升级测试。当前 main 对正常 fixture 成功，并不能据此宣告 #154 修复。下一项最有价值的新增证据是**实际进程的 `tlsmlkem` 设置和一个带 TLS 完成事件的同环境 GET trace**。
-
-## 复现实验
-
-从本任务 worktree 根目录执行；工具均使用本地源码，输出目录必须是新建临时目录。
-
-```sh
-LAB_DIR=$(mktemp -d)
-CGO_ENABLED=0 GOWORK=off go build -mod=readonly -tags kqueue \
-  -o "$LAB_DIR/silo" .
-CGO_ENABLED=0 GOWORK=off go build -mod=readonly -tags kqueue \
-  -o "$LAB_DIR/probe" docs/investigations/issue-154/probe.go
-go build -o "$LAB_DIR/fixture" docs/investigations/issue-154/fixture.go
-go build -mod=readonly -o "$LAB_DIR/admin-check" docs/investigations/issue-154/admin-check.go
-"$LAB_DIR/fixture" -dir "$LAB_DIR/idp" > "$LAB_DIR/hello.jsonl" 2> "$LAB_DIR/fixture.log" &
-FIXTURE_PID=$!
-trap 'kill "$FIXTURE_PID" 2>/dev/null || true' EXIT
-attempt=0
-while [ ! -s "$LAB_DIR/idp/url" ] && [ "$attempt" -lt 50 ]; do
-  sleep 0.1
-  attempt=$((attempt + 1))
-done
-test -s "$LAB_DIR/idp/url" || exit 1
-# 私钥仅保存在 fixture 内存。
-OIDC_URL="$(cat "$LAB_DIR/idp/url")/.well-known/openid-configuration"
-export OIDC_URL
-"$LAB_DIR/probe" -ca "$LAB_DIR/idp/ca.pem"
-printf '%s' reject-mlkem > "$LAB_DIR/idp/mode"
-GODEBUG=tlsmlkem=0 "$LAB_DIR/probe" -ca "$LAB_DIR/idp/ca.pem"
-GODEBUG=tlsmlkem=0 "$LAB_DIR/probe" -ca "$LAB_DIR/idp/ca.pem" -default-curves
-printf '%s' reject-mldsa > "$LAB_DIR/idp/mode"
-"$LAB_DIR/probe" -ca "$LAB_DIR/idp/ca.pem" -classical
-"$LAB_DIR/probe" -ca "$LAB_DIR/idp/ca.pem" -tls12
-kill "$FIXTURE_PID"
-```
-
-完整 Server 实验使用上述临时目录下的 `certs/CAs` 和数据目录，设置虚构的 `MINIO_IDENTITY_OPENID_CLIENT_ID`、fixture URL、临时 root 凭据，并显式指定 `--config-dir`、`--certs-dir`、loopback API/Console 地址。用 mode 文件切换 `reset`/`bad-jwks` 到 `normal`，同时检查 cluster health 和 Admin ListUsers；结果已保存在 `evidence.json`。`admin-check add` 只操作 `LAB_SERVER=127.0.0.1:port` 的实验实例，读取临时 `LAB_USER`、`LAB_PASSWORD`、`LAB_OIDC_URL`，使用虚构 OIDC secret。
-
-跨版本构建时从各 tag `git archive` 获取源码，把 `probe.go` 放入该 module 根目录；旧版探针仅将 certs 导入替换为 `github.com/minio/pkg/v3/certs`，由旧版 go.mod 选择 PGSTY v3.11.0。分别强制 `GOTOOLCHAIN=go1.26.5` / `go1.27.1`，使用 `-mod=readonly`，不修改历史 go.mod。另将旧源码原依赖用 Go 1.27.1 构建，隔离工具链因素。若为 Linux 客户端构建探针，按已确认的架构设置 `GOOS=linux GOARCH=amd64` 或 `arm64`，不使用 Server 镜像代替。
-
-### 第二轮 Linux 完整 Server 对照
-
-从本任务根目录运行以下 Bash 命令。需要预先准备 Go 1.26.5、1.27.1 及源码依赖缓存，以及带 Python 3、curl/OpenSSL/HTTP2 的通用 Linux 基础镜像。`GOTOOLCHAIN` 明确选定编译器，`-mod=readonly` 保留依赖版本；实际测试构建用已安装工具链的绝对路径配合 `GOTOOLCHAIN=local`，等价地避免自动切换编译器。以下固定 `arm64` 与本次实验一致。
-
-```bash
-LAB_DIR=$(mktemp -d)
-mkdir -p "$LAB_DIR/old" "$LAB_DIR/head" "$LAB_DIR/candidate" "$LAB_DIR/bin" "$LAB_DIR/out"
-git archive d88f46ccee345a9c2fabe2d221d9a9e56bc11aec | tar -x -C "$LAB_DIR/old"
-git archive d1105bbb3d4a0afa33b3a4ac11b821235038ed0e | tar -x -C "$LAB_DIR/head"
-git archive d1105bbb3d4a0afa33b3a4ac11b821235038ed0e | tar -x -C "$LAB_DIR/candidate"
-git -C "$LAB_DIR/candidate" apply "$PWD/docs/investigations/issue-154/openid-default-curves.patch"
-
-(cd "$LAB_DIR/old" && CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 \
-  GOTOOLCHAIN=go1.26.5 go build -mod=readonly -tags kqueue -o "$LAB_DIR/bin/old-go126" .)
-(cd "$LAB_DIR/old" && CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 \
-  GOTOOLCHAIN=go1.27.1 go build -mod=readonly -tags kqueue -o "$LAB_DIR/bin/old-go127" .)
-(cd "$LAB_DIR/head" && CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 \
-  GOTOOLCHAIN=go1.27.1 go build -mod=readonly -tags kqueue -o "$LAB_DIR/bin/head-go127" .)
-(cd "$LAB_DIR/candidate" && CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 \
-  GOTOOLCHAIN=go1.27.1 go build -mod=readonly -tags kqueue -o "$LAB_DIR/bin/candidate-go127" .)
-CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 GOTOOLCHAIN=go1.27.1 \
-  go build -mod=readonly -o "$LAB_DIR/bin/fixture" docs/investigations/issue-154/fixture.go
-CGO_ENABLED=0 GOWORK=off GOOS=linux GOARCH=arm64 GOTOOLCHAIN=go1.27.1 \
-  go build -mod=readonly -o "$LAB_DIR/bin/admin-check" docs/investigations/issue-154/admin-check.go
-
-# 本机已存在的通用 Debian 12 arm64 基础镜像；不拉取镜像，不映射端口。
-LINUX_BASE_IMAGE=sha256:307af7711e2e04ab75759cb42a1eef45c43c4404894c0e30dd19f742b107b922
-docker run --rm --pull=never --network none \
-  --mount "type=bind,source=$LAB_DIR/bin,target=/lab/bin,readonly" \
-  --mount "type=bind,source=$LAB_DIR/out,target=/lab/out" \
-  --mount "type=bind,source=$PWD/docs/investigations/issue-154/run-linux.py,target=/lab/run-linux.py,readonly" \
-  --entrypoint python3 "$LINUX_BASE_IMAGE" /lab/run-linux.py
-```
-
-每组场景输出一行摘要，同时在独立输出目录保存 `result.json`。脚本用 `finally` 终止其 Server/fixture，容器结束自动删除。它没有导出会话 cookie、授权码、JWT、state 或临时密码；原始运行目录只用于该次隔离实验，交付证据只保留握手、结果和构建身份。
@@ -1,41 +0,0 @@
-//go:build ignore
-
-// Loopback-only lab client for the Admin API used by Console's OIDC form.
-package main
-
-import (
-	"context"
-	"fmt"
-	"net"
-	"os"
-	"time"
-
-	"github.com/minio/madmin-go/v3"
-)
-
-func main() {
-	endpoint := os.Getenv("LAB_SERVER")
-	host, _, err := net.SplitHostPort(endpoint)
-	if err != nil || host != "127.0.0.1" {
-		panic("LAB_SERVER must use IPv4 loopback")
-	}
-	a, err := madmin.New(endpoint, os.Getenv("LAB_USER"), os.Getenv("LAB_PASSWORD"), false)
-	if err != nil {
-		panic(err)
-	}
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-	defer cancel()
-	if len(os.Args) > 1 && os.Args[1] == "add" {
-		restart, err := a.AddOrUpdateIDPConfig(ctx, "openid", "local154", "enable=on client_id=local154 client_secret=local154-placeholder config_url="+os.Getenv("LAB_OIDC_URL"), false)
-		fmt.Printf("add restart=%v err=%v\n", restart, err)
-		if err != nil {
-			os.Exit(1)
-		}
-		return
-	}
-	users, err := a.ListUsers(ctx)
-	fmt.Printf("list_users count=%d err=%v\n", len(users), err)
-	if err != nil {
-		os.Exit(1)
-	}
-}
@@ -1,41 +0,0 @@
-//go:build ignore
-
-// Run only with the public, synthetic CA made by fixture.go.
-package main
-
-import (
-	"crypto/x509"
-	"encoding/json"
-	"encoding/pem"
-	"os"
-	"runtime"
-
-	"github.com/pgsty/silo-pkg/v3/certs"
-)
-
-func main() {
-	if len(os.Args) != 3 {
-		panic("usage: cert-roots synthetic-ca.pem explicit-ca-path-or-empty")
-	}
-	data, err := os.ReadFile(os.Args[1])
-	if err != nil {
-		panic(err)
-	}
-	block, _ := pem.Decode(data)
-	if block == nil || block.Type != "CERTIFICATE" {
-		panic("expected public certificate")
-	}
-	certificate, err := x509.ParseCertificate(block.Bytes)
-	if err != nil {
-		panic(err)
-	}
-	roots, err := certs.GetRootCAs(os.Args[2])
-	if err != nil {
-		panic(err)
-	}
-	_, err = certificate.Verify(x509.VerifyOptions{Roots: roots})
-	_ = json.NewEncoder(os.Stdout).Encode(map[string]any{
-		"go": runtime.Version(), "os": runtime.GOOS,
-		"explicit_ca": os.Args[2] != "", "trusted": err == nil,
-	})
-}
@@ -1,179 +0,0 @@
-//go:build ignore
-
-// Loopback-only synthetic OIDC/TLS fixture. Only disposable lab identities are used.
-// Rejection modes model hypotheses; they are not evidence about the user's IdP.
-package main
-
-import (
-	"crypto"
-	"crypto/rand"
-	"crypto/rsa"
-	"crypto/sha256"
-	"crypto/tls"
-	"crypto/x509"
-	"crypto/x509/pkix"
-	"encoding/base64"
-	"encoding/json"
-	"encoding/pem"
-	"errors"
-	"flag"
-	"fmt"
-	"math/big"
-	"net"
-	"net/http"
-	"net/url"
-	"os"
-	"path/filepath"
-	"slices"
-	"strings"
-	"sync"
-	"time"
-)
-
-type observedConn struct {
-	net.Conn
-	readBytes int
-}
-
-func (c *observedConn) Read(p []byte) (int, error) {
-	n, e := c.Conn.Read(p)
-	c.readBytes += n
-	return n, e
-}
-func (c *observedConn) reset() { _ = c.Conn.(*net.TCPConn).SetLinger(0); _ = c.Conn.Close() }
-
-type observedListener struct{ net.Listener }
-
-func (l observedListener) Accept() (net.Conn, error) {
-	c, e := l.Listener.Accept()
-	if e != nil {
-		return nil, e
-	}
-	return &observedConn{Conn: c}, nil
-}
-
-func main() {
-	dir := flag.String("dir", "", "isolated output directory for public CA, URL, and mode file")
-	tls13 := flag.Bool("tls13", false, "allow TLS 1.3 in addition to the TLS 1.2 baseline")
-	flag.Parse()
-	if *dir == "" {
-		panic("-dir required")
-	}
-	must(os.MkdirAll(*dir, 0700))
-	key, err := rsa.GenerateKey(rand.Reader, 2048)
-	must(err)
-	root := &x509.Certificate{SerialNumber: big.NewInt(1), Subject: pkix.Name{CommonName: "issue-154 local CA"}, NotBefore: time.Now().Add(-time.Hour), NotAfter: time.Now().Add(24 * time.Hour), IsCA: true, BasicConstraintsValid: true, KeyUsage: x509.KeyUsageCertSign | x509.KeyUsageDigitalSignature}
-	rootDER, err := x509.CreateCertificate(rand.Reader, root, root, &key.PublicKey, key)
-	must(err)
-	leaf := &x509.Certificate{SerialNumber: big.NewInt(2), Subject: pkix.Name{CommonName: "localhost"}, DNSNames: []string{"localhost"}, IPAddresses: []net.IP{net.ParseIP("127.0.0.1")}, NotBefore: root.NotBefore, NotAfter: root.NotAfter, ExtKeyUsage: []x509.ExtKeyUsage{x509.ExtKeyUsageServerAuth}, KeyUsage: x509.KeyUsageDigitalSignature}
-	leafDER, err := x509.CreateCertificate(rand.Reader, leaf, root, &key.PublicKey, key)
-	must(err)
-	must(os.WriteFile(filepath.Join(*dir, "ca.pem"), pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: rootDER}), 0600))
-	ln, err := net.Listen("tcp", "127.0.0.1:0")
-	must(err)
-	base := "https://" + ln.Addr().String()
-	must(os.WriteFile(filepath.Join(*dir, "url"), []byte(base), 0600))
-	mode := func() string { b, _ := os.ReadFile(filepath.Join(*dir, "mode")); return strings.TrimSpace(string(b)) }
-	var mu sync.Mutex
-	log := func(v any) { mu.Lock(); defer mu.Unlock(); _ = json.NewEncoder(os.Stdout).Encode(v) }
-	tc := &tls.Config{Certificates: []tls.Certificate{{Certificate: [][]byte{leafDER, rootDER}, PrivateKey: key}}, MinVersion: tls.VersionTLS12, MaxVersion: tls.VersionTLS12, CurvePreferences: []tls.CurveID{tls.CurveP256}, CipherSuites: []uint16{tls.TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384}}
-	if *tls13 {
-		tc.MaxVersion = tls.VersionTLS13
-	}
-	peerConfig := tc.Clone()
-	peerConfig.NextProtos = []string{"h2", "http/1.1"}
-	// net/http validates HTTP/2 support before GetConfigForClient; the fixture
-	// then deliberately selects only the reported AES-256 suite.
-	tc.CipherSuites = append(tc.CipherSuites, tls.TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256)
-	tc.GetConfigForClient = func(chi *tls.ClientHelloInfo) (*tls.Config, error) {
-		m := mode()
-		c := chi.Conn.(*observedConn)
-		log(map[string]any{"event": "hello", "mode": m, "bytes_read": c.readBytes, "curves": chi.SupportedCurves, "signatures": chi.SignatureSchemes, "alpn": chi.SupportedProtos, "versions": chi.SupportedVersions})
-		reject := m == "reset" || m == "reject-mlkem" && slices.Contains(chi.SupportedCurves, tls.CurveID(4588)) || m == "reject-mldsa" && slices.Contains(chi.SignatureSchemes, tls.SignatureScheme(0x0904)) || m == "require-h2" && !slices.Contains(chi.SupportedProtos, "h2")
-		if reject {
-			c.reset()
-			return nil, errors.New("synthetic ClientHello rejection")
-		}
-		return peerConfig, nil
-	}
-	server := &http.Server{TLSConfig: tc, ReadHeaderTimeout: 5 * time.Second}
-	var codes sync.Map
-	server.Handler = http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		log(map[string]any{"event": "request", "path": r.URL.Path, "protocol": r.Proto, "tls": r.TLS.Version, "cipher": r.TLS.CipherSuite, "resumed": r.TLS.DidResume, "ua": r.UserAgent()})
-		if mode() == "reject-silo-ua" && strings.HasPrefix(r.UserAgent(), "Silo") {
-			c, _, e := w.(http.Hijacker).Hijack()
-			if e == nil {
-				c.(*tls.Conn).NetConn().(*observedConn).reset()
-			}
-			return
-		}
-		w.Header().Set("Content-Type", "application/json")
-		switch r.URL.Path {
-		case "/authorize":
-			q := r.URL.Query()
-			redirect, err := url.Parse(q.Get("redirect_uri"))
-			if err != nil || redirect.Scheme != "http" || redirect.Hostname() != "127.0.0.1" || redirect.Path != "/oauth_callback" || q.Get("client_id") != "local154" {
-				http.Error(w, "loopback lab authorization only", 400)
-				return
-			}
-			codeBytes := make([]byte, 18)
-			_, err = rand.Read(codeBytes)
-			must(err)
-			code := base64.RawURLEncoding.EncodeToString(codeBytes)
-			codes.Store(code, q.Get("nonce"))
-			values := redirect.Query()
-			values.Set("code", code)
-			values.Set("state", q.Get("state"))
-			redirect.RawQuery = values.Encode()
-			http.Redirect(w, r, redirect.String(), http.StatusFound)
-		case "/token":
-			if r.Method != http.MethodPost || r.ParseForm() != nil {
-				http.Error(w, "bad token request", 400)
-				return
-			}
-			id, secret, ok := r.BasicAuth()
-			if !ok {
-				id, secret = r.Form.Get("client_id"), r.Form.Get("client_secret")
-			}
-			nonce, found := codes.LoadAndDelete(r.Form.Get("code"))
-			if id != "local154" || secret != "local154-placeholder" || !found || r.Form.Get("grant_type") != "authorization_code" {
-				w.WriteHeader(400)
-				_ = json.NewEncoder(w).Encode(map[string]string{"error": "invalid_grant"})
-				return
-			}
-			audience := id
-			if mode() == "bad-audience" {
-				audience = "different-lab-client"
-			}
-			claims, _ := json.Marshal(map[string]any{"iss": base, "sub": "local154-user", "aud": audience, "iat": time.Now().Unix(), "exp": time.Now().Add(time.Hour).Unix(), "policy": "readwrite", "nonce": nonce})
-			header := base64.RawURLEncoding.EncodeToString([]byte(`{"alg":"RS256","kid":"local-154","typ":"JWT"}`))
-			payload := header + "." + base64.RawURLEncoding.EncodeToString(claims)
-			hash := sha256.Sum256([]byte(payload))
-			signature, err := rsa.SignPKCS1v15(rand.Reader, key, crypto.SHA256, hash[:])
-			must(err)
-			if mode() == "bad-signature" {
-				signature[0] ^= 1
-			}
-			token := payload + "." + base64.RawURLEncoding.EncodeToString(signature)
-			_ = json.NewEncoder(w).Encode(map[string]any{"access_token": token, "id_token": token, "token_type": "Bearer", "expires_in": 3600})
-		case "/.well-known/openid-configuration":
-			_ = json.NewEncoder(w).Encode(map[string]any{"issuer": base, "jwks_uri": base + "/jwks", "authorization_endpoint": base + "/authorize", "token_endpoint": base + "/token", "response_types_supported": []string{"code"}, "subject_types_supported": []string{"public"}, "id_token_signing_alg_values_supported": []string{"RS256"}, "scopes_supported": []string{"openid"}})
-		case "/jwks":
-			if mode() == "bad-jwks" {
-				http.Error(w, "synthetic JWKS outage", 503)
-				return
-			}
-			_ = json.NewEncoder(w).Encode(map[string]any{"keys": []any{map[string]any{"kty": "RSA", "kid": "local-154", "use": "sig", "alg": "RS256", "n": base64.RawURLEncoding.EncodeToString(key.N.Bytes()), "e": "AQAB"}}})
-		default:
-			http.NotFound(w, r)
-		}
-	})
-	fmt.Fprintln(os.Stderr, base)
-	must(server.ServeTLS(observedListener{ln}, "", ""))
-}
-
-func must(err error) {
-	if err != nil {
-		panic(err)
-	}
-}
@@ -1,40 +0,0 @@
-# Historical OIDC-only candidate; superseded by ../go127-stack.md. Do not apply on top of the stack fix.
--- a/cmd/utils.go
-+++ b/cmd/utils.go
-@@ -654,6 +654,14 @@
- 	return NewHTTPTransportWithTimeout(1 * time.Minute)
- }
- 
-+// NewOpenIDHTTPTransport uses Go defaults for external identity-provider key exchange.
-+// This lets tlsmlkem/tlssecpmlkem configure their documented default sets.
-+func NewOpenIDHTTPTransport() *http.Transport {
-+	tr := NewHTTPTransport()
-+	tr.TLSClientConfig.CurvePreferences = nil
-+	return tr
-+}
-+
- // Default values for dial timeout
- const defaultDialTimeout = 5 * time.Second
- 
--- a/cmd/iam.go
-+++ b/cmd/iam.go
-@@ -277,7 +277,7 @@
- 	for {
- 		if !openidInit {
- 			openidConfig, err := openid.LookupConfig(s,
-				xhttp.WithUserAgent(NewHTTPTransport(), func() string {
-+				xhttp.WithUserAgent(NewOpenIDHTTPTransport(), func() string {
- 					return getUserAgent(getMinioMode())
- 				}), xhttp.DrainBody, globalSite.Region())
- 			if err != nil {
--- a/cmd/config-current.go
-+++ b/cmd/config-current.go
-@@ -352,7 +352,7 @@
- 		}
- 	case config.IdentityOpenIDSubSys:
- 		if _, err := openid.LookupConfig(s,
-			xhttp.WithUserAgent(NewHTTPTransport(), func() string {
-+			xhttp.WithUserAgent(NewOpenIDHTTPTransport(), func() string {
- 				return getUserAgent(getMinioMode())
- 			}), xhttp.DrainBody, globalSite.Region()); err != nil {
- 			return err
@@ -1,165 +0,0 @@
-//go:build ignore
-
-// Diagnostic GET using the Server's actual transport constructor.
-// Build explicitly from the SILO module root; see ../issue-154.md.
-package main
-
-import (
-	"context"
-	"crypto/tls"
-	"errors"
-	"flag"
-	"fmt"
-	"io"
-	"net"
-	"net/http"
-	"net/http/httptrace"
-	"net/url"
-	"os"
-	"runtime"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/minio/minio/cmd"
-	xhttp "github.com/minio/minio/internal/http"
-	"github.com/pgsty/silo-pkg/v3/certs"
-)
-
-func main() {
-	endpoint := flag.String("url", os.Getenv("OIDC_URL"), "discovery URL; no credentials or query string")
-	ca := flag.String("ca", "", "same CA file or certs/CAs directory as Server")
-	h2 := flag.Bool("h2", false, "diagnostic: opt in to HTTP/2")
-	classical := flag.Bool("classical", false, "diagnostic: omit hybrid key exchange only")
-	defaultCurves := flag.Bool("default-curves", false, "diagnostic: let Go choose curves and honor its GODEBUG defaults")
-	tls12 := flag.Bool("tls12", false, "diagnostic: TLS 1.2 only; keeps certificate verification")
-	direct := flag.Bool("direct", false, "diagnostic: bypass environment proxy")
-	ip := flag.String("ip", "", "diagnostic: pin destination IP, preserving Host/SNI; requires -direct")
-	fresh := flag.Bool("fresh", false, "diagnostic: close idle connections between requests")
-	ua := flag.String("ua", "issue-154-probe", "HTTP User-Agent; supply actual Server UA to investigate a WAF")
-	n := flag.Int("n", 1, "number of GETs (1 to 3)")
-	flag.Parse()
-	u, err := url.Parse(*endpoint)
-	if err != nil || u.Scheme != "https" || u.Host == "" || u.User != nil || u.RawQuery != "" || u.Fragment != "" || *n < 1 || *n > 3 {
-		fmt.Fprintln(os.Stderr, "require an HTTPS URL without credentials/query/fragment and -n between 1 and 3")
-		os.Exit(2)
-	}
-	if *ip != "" && (!*direct || net.ParseIP(*ip) == nil) {
-		fmt.Fprintln(os.Stderr, "-ip requires a literal IP and -direct")
-		os.Exit(2)
-	}
-	if *classical && *defaultCurves {
-		fmt.Fprintln(os.Stderr, "choose at most one of -classical and -default-curves")
-		os.Exit(2)
-	}
-	tr := cmd.NewHTTPTransport()
-	tr.TLSClientConfig.RootCAs, err = certs.GetRootCAs(*ca)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "CA loading failed; check the local CA path")
-		os.Exit(2)
-	}
-	if *h2 {
-		tr.ForceAttemptHTTP2 = true
-	}
-	if *classical {
-		tr.TLSClientConfig.CurvePreferences = []tls.CurveID{tls.CurveP256, tls.X25519, tls.CurveP384, tls.CurveP521}
-	}
-	if *defaultCurves {
-		tr.TLSClientConfig.CurvePreferences = nil
-	}
-	if *tls12 {
-		tr.TLSClientConfig.MinVersion = tls.VersionTLS12
-		tr.TLSClientConfig.MaxVersion = tls.VersionTLS12
-	}
-	if *direct {
-		tr.Proxy = nil
-	}
-	if *ip != "" {
-		base := tr.DialContext
-		tr.DialContext = func(ctx context.Context, network, address string) (net.Conn, error) {
-			host, port, e := net.SplitHostPort(address)
-			if e != nil {
-				return nil, e
-			}
-			if host == u.Hostname() {
-				address = net.JoinHostPort(*ip, port)
-			}
-			return base(ctx, network, address)
-		}
-	}
-	defer tr.CloseIdleConnections()
-	var mu sync.Mutex
-	log := func(format string, args ...any) { mu.Lock(); defer mu.Unlock(); fmt.Printf(format+"\n", args...) }
-	log("go=%s os=%s arch=%s h2=%v classical=%v default_curves=%v tls12=%v", runtime.Version(), runtime.GOOS, runtime.GOARCH, *h2, *classical, *defaultCurves, *tls12)
-	// Deliberately print no URL, headers, body, client ID, secret, or token.
-	req, _ := http.NewRequest(http.MethodGet, u.String(), nil)
-	proxy := "direct"
-	if tr.Proxy != nil {
-		p, e := tr.Proxy(req)
-		if e != nil {
-			log("proxy_selection_error=%T", e)
-			os.Exit(2)
-		}
-		if p != nil {
-			proxy = p.Scheme + " proxy (address omitted)"
-		}
-	}
-	log("route=%s curves=%v", proxy, tr.TLSClientConfig.CurvePreferences)
-	client := &http.Client{Transport: xhttp.WithUserAgent(tr, func() string { return *ua }), Timeout: 20 * time.Second,
-		CheckRedirect: func(_ *http.Request, _ []*http.Request) error { return http.ErrUseLastResponse }}
-	failed := false
-	for i := 0; i < *n; i++ {
-		if *fresh {
-			tr.CloseIdleConnections()
-		}
-		log("request=%d", i+1)
-		trace := &httptrace.ClientTrace{
-			DNSDone:           func(d httptrace.DNSDoneInfo) { log("dns_addresses=%v err=%s", d.Addrs, errorClass(d.Err)) },
-			ConnectStart:      func(network, addr string) { log("connect=%s %s", network, addr) },
-			ConnectDone:       func(_, addr string, e error) { log("connected=%s err=%s", addr, errorClass(e)) },
-			TLSHandshakeStart: func() { log("tls_start") },
-			TLSHandshakeDone: func(s tls.ConnectionState, e error) {
-				log("tls_done=0x%x cipher=%s alpn=%q resumed=%v verified_chains=%d err=%s", s.Version, tls.CipherSuiteName(s.CipherSuite), s.NegotiatedProtocol, s.DidResume, len(s.VerifiedChains), errorClass(e))
-			},
-			GotConn:              func(c httptrace.GotConnInfo) { log("got_conn=%s reused=%v", c.Conn.RemoteAddr(), c.Reused) },
-			WroteRequest:         func(w httptrace.WroteRequestInfo) { log("wrote_request err=%s", errorClass(w.Err)) },
-			GotFirstResponseByte: func() { log("first_response_byte") },
-		}
-		r := req.Clone(httptrace.WithClientTrace(context.Background(), trace))
-		resp, e := client.Do(r)
-		if e != nil {
-			log("get_error=%s", errorClass(e))
-			failed = true
-			continue
-		}
-		log("status=%d protocol=%s", resp.StatusCode, resp.Proto)
-		_, e = io.Copy(io.Discard, io.LimitReader(resp.Body, 1<<20))
-		resp.Body.Close()
-		if e != nil || resp.StatusCode != http.StatusOK {
-			failed = true
-			log("body_error=%s", errorClass(e))
-		}
-	}
-	if failed {
-		os.Exit(1)
-	}
-}
-
-func errorClass(err error) string {
-	if err == nil {
-		return "none"
-	}
-	if errors.Is(err, context.DeadlineExceeded) {
-		return "deadline"
-	}
-	// Error text may contain a private URL. Emit only category and concrete type.
-	category := "other"
-	s := err.Error()
-	for _, k := range []string{"connection reset by peer", "x509:", "TLS handshake timeout", "connection refused", "EOF"} {
-		if strings.Contains(s, k) {
-			category = k
-			break
-		}
-	}
-	return fmt.Sprintf("%s (%T)", category, err)
-}
@@ -1,170 +0,0 @@
-#!/usr/bin/env python3
-"""Bounded, loopback-only full-Server comparison. See the investigation report.
-
-Run in an isolated generic Linux container with locally built binaries in
-/lab/bin and a new disposable /lab/out. No customer identities or endpoints.
-"""
-import http.cookiejar
-import json
-import os
-from pathlib import Path
-import secrets
-import shutil
-import socket
-import ssl
-import subprocess
-import time
-import urllib.error
-import urllib.parse
-import urllib.request
-
-ROOT = Path("/lab")
-OUT = ROOT / "out"
-BASE = {k: v for k, v in os.environ.items()
-        if not k.startswith(("MINIO_", "SILO_", "CONSOLE_"))
-        and k.lower() not in {"http_proxy", "https_proxy", "all_proxy", "no_proxy", "godebug"}}
-
-
-def port():
-    with socket.socket() as sock:
-        sock.bind(("127.0.0.1", 0))
-        return sock.getsockname()[1]
-
-
-def request(url, opener=None, payload=None):
-    headers = {"Origin": f"http://{urllib.parse.urlparse(url).netloc}"}
-    if payload is not None:
-        headers["Content-Type"] = "application/json"
-    req = urllib.request.Request(url, data=None if payload is None else json.dumps(payload).encode(), headers=headers)
-    try:
-        response = (opener.open if opener else urllib.request.urlopen)(req, timeout=2)
-    except urllib.error.HTTPError as err:
-        response = err
-    except (urllib.error.URLError, TimeoutError):
-        return 0, {}, b""
-    with response:
-        return response.code, dict(response.headers), response.read(1 << 20)
-
-
-def stop(proc):
-    proc.terminate()
-    try:
-        proc.wait(timeout=4)
-    except subprocess.TimeoutExpired:
-        proc.kill()
-        proc.wait(timeout=2)
-
-
-class NoRedirect(urllib.request.HTTPRedirectHandler):
-    def redirect_request(self, req, fp, code, msg, headers, newurl):
-        return None
-
-
-def login(console, ca):
-    jar = http.cookiejar.CookieJar()
-    client = urllib.request.build_opener(urllib.request.HTTPCookieProcessor(jar))
-    status, _, raw = request(console + "/api/v1/login", client)
-    details = json.loads(raw)
-    rules = details.get("redirectRules", [])
-    assert status == 200 and len(rules) == 1, (status, details)
-    auth_url = rules[0]["redirect"]
-    provider = urllib.request.build_opener(NoRedirect(), urllib.request.HTTPSHandler(context=ssl.create_default_context(cafile=str(ca))))
-    status, headers, _ = request(auth_url, provider)
-    callback = headers.get("Location", headers.get("location", ""))
-    assert status == 302 and callback.startswith(console + "/oauth_callback?"), (status, callback)
-    values = urllib.parse.parse_qs(urllib.parse.urlparse(callback).query)
-    callback_status, _, _ = request(callback, client)
-    status, _, _ = request(console + "/api/v1/login/oauth2/auth", client,
-                           {"code": values["code"][0], "state": values["state"][0]})
-    buckets_status, _, _ = request(console + "/api/v1/buckets", client)
-    # Do not record cookies, codes, JWTs, or the state value.
-    return {"callback_status": callback_status, "login_status": status,
-            "buckets_status": buckets_status, "session_cookie": any(c.name == "token" for c in jar)}
-
-
-def run(name, binary, mode="normal", debug=None, tls13=False,
-        expected=True, trusted=True, oidc=True, oauth=False, add=False):
-    d = OUT / name
-    (d / "certs/CAs").mkdir(parents=True, exist_ok=False)
-    idp = d / "idp"
-    idp.mkdir()
-    (idp / "mode").write_text(mode)
-    with (d / "fixture.jsonl").open("w") as events, (d / "fixture.stderr").open("w") as errors, (d / "server.log").open("w") as logs:
-        fixture = subprocess.Popen([str(ROOT / "bin/fixture"), "-dir", str(idp), *(["-tls13"] if tls13 else [])], env=BASE, stdout=events, stderr=errors)
-        server = None
-        try:
-            until = time.monotonic() + 5
-            while not (idp / "url").is_file() and time.monotonic() < until:
-                time.sleep(.05)
-            assert (idp / "url").is_file(), "fixture did not initialize"
-            url = (idp / "url").read_text() + "/.well-known/openid-configuration"
-            if trusted:
-                shutil.copyfile(idp / "ca.pem", d / "certs/CAs/lab.pem")
-            sport, cport = port(), port()
-            address = f"127.0.0.1:{sport}"
-            api, console = "http://" + address, f"http://127.0.0.1:{cport}"
-            password = secrets.token_urlsafe(24)
-            env = dict(BASE, MINIO_ROOT_USER="local154", MINIO_ROOT_PASSWORD=password, MINIO_BROWSER="on")
-            if debug:
-                env["GODEBUG"] = debug
-            if oidc:
-                env.update(MINIO_IDENTITY_OPENID_CONFIG_URL=url,
-                           MINIO_IDENTITY_OPENID_CLIENT_ID="local154",
-                           MINIO_IDENTITY_OPENID_CLIENT_SECRET="local154-placeholder",
-                           MINIO_IDENTITY_OPENID_REDIRECT_URI=console + "/oauth_callback")
-            server = subprocess.Popen([str(ROOT / "bin" / binary), "--config-dir", str(d / "config"), "--certs-dir", str(d / "certs"), "server", "--address", address, "--console-address", f"127.0.0.1:{cport}", str(d / "data")], env=env, stdout=logs, stderr=subprocess.STDOUT)
-            until = time.monotonic() + 15
-            while time.monotonic() < until:
-                assert server.poll() is None, "Server exited; inspect its local log"
-                status, _, _ = request(api + "/minio/health/cluster")
-                if status == 200 and request(console)[0] == 200:
-                    break
-                if not expected and (d / "server.log").read_text().count("Waiting for OpenID") >= 2:
-                    break
-                time.sleep(.1)
-            result = {"case": name, "binary": binary, "mode": mode, "godebug": debug, "tls13": tls13,
-                      "cluster": request(api + "/minio/health/cluster")[0],
-                      "ready": request(api + "/minio/health/ready")[0], "console": request(console)[0]}
-            assert result["cluster"] == (200 if expected else 503), result
-            aenv = dict(BASE, LAB_SERVER=address, LAB_USER="local154", LAB_PASSWORD=password, LAB_OIDC_URL=url)
-            if expected:
-                admin = subprocess.run([str(ROOT / "bin/admin-check")], env=aenv, capture_output=True, text=True, timeout=7)
-                result["admin_list_ok"] = admin.returncode == 0
-                assert result["admin_list_ok"], admin.stdout
-            curl = subprocess.run(["curl", "--cacert", str(idp / "ca.pem"), "--http2", "--max-time", "3", "-sS", "-o", "/dev/null", "-w", "%{http_code} %{http_version}", url], env=BASE, capture_output=True, text=True, timeout=5)
-            result["curl"] = {"exit": curl.returncode, "status_protocol": curl.stdout}
-            if add:
-                attempt = subprocess.run([str(ROOT / "bin/admin-check"), "add"], env=aenv, capture_output=True, text=True, timeout=7)
-                result["add_ok"] = attempt.returncode == 0
-                result["add_reset"] = "connection reset by peer" in attempt.stdout
-                assert result["add_ok"] == binary.startswith("candidate"), result
-            if oauth:
-                result["oauth"] = login(console, idp / "ca.pem")
-                assert result["oauth"]["login_status"] == 204 and result["oauth"]["buckets_status"] == 200, result
-                for bad in ("bad-signature", "bad-audience"):
-                    (idp / "mode").write_text(bad)
-                    result[bad] = login(console, idp / "ca.pem")
-                    assert result[bad]["login_status"] >= 400 and result[bad]["buckets_status"] >= 400, result
-            # Public handshake metadata only; no authorization parameters.
-            result["events"] = [json.loads(line) for line in (d / "fixture.jsonl").read_text().splitlines()]
-            (d / "result.json").write_text(json.dumps(result, indent=2) + "\n")
-            print(json.dumps({k: v for k, v in result.items() if k != "events"}), flush=True)
-        finally:
-            if server is not None:
-                stop(server)
-            stop(fixture)
-
-
-if __name__ == "__main__":
-    run("old126-normal", "old-go126")
-    run("old127-normal", "old-go127")
-    run("old126-compat", "old-go126", "reject-mlkem", "tlsmlkem=0")
-    run("old127-compat", "old-go127", "reject-mlkem", "tlsmlkem=0", expected=False)
-    run("head127-compat", "head-go127", "reject-mlkem", "tlsmlkem=0", expected=False)
-    run("candidate127-compat-login", "candidate-go127", "reject-mlkem", "tlsmlkem=0", oauth=True)
-    run("candidate127-no-optout", "candidate-go127", "reject-mlkem", expected=False)
-    run("candidate127-tls13-login", "candidate-go127", tls13=True, oauth=True)
-    run("candidate127-untrusted", "candidate-go127", trusted=False, expected=False)
-    run("candidate127-mldsa", "candidate-go127", "reject-mldsa", "tlsmlkem=0", expected=False)
-    run("head127-add", "head-go127", "reject-mlkem", "tlsmlkem=0", oidc=False, add=True)
-    run("candidate127-add", "candidate-go127", "reject-mlkem", "tlsmlkem=0", oidc=False, add=True)
@@ -1,103 +0,0 @@
-# Issue #77 当前核验与最小修复建议
-
-> 历史快照（2026-09-12，实施前）。其中的 Issue 状态、待办和方案约束只描述当时情况；最终实现、修正与验收以 [归档总记录](issue-77.md) 为准。
-
-核验日期：2026-09-12。对象：[SILO #77](https://github.com/pgsty/silo/issues/77)。代码基准：远端 `main` 的 `5c576581631561c446f30ae5b566f0aa793adc1c`，在独立 detached worktree 中运行测试。用户工作目录仍位于 `12f631b50`；两者差异为 #179 的 federation 修复，不涉及本次复制元数据代码。
-
-**结论：问题真实，当前主干仍未修完，应保持打开。原方案的 A/B 核心必要，但不能照搬 8 月的实施清单；已有基础设施可以复用，同时必须补上批量复制入口和相同内容的时间戳同步。**
-
-本次是分析与复现，没有修改产品代码、提交、发布或修改 GitHub Issue。下述测试是本地 ObjectLayer 与进程内 HTTP/RPC 复现，不是线上多站点验收。
-
-## 1. 当前状态
-
-GitHub API 实时返回 #77 为 `OPEN`，未分配负责人或 milestone，最后更新为 `2026-09-08T14:45:10Z`。核验时仓库没有打开的 PR。
-
-| 范围 | 当前事实 | 是否仍属遗留项 |
-| --- | --- | --- |
-| #77-C，各站点统计 | [#91](https://github.com/pgsty/silo/pull/91) 于 8 月 29 日合并；本次现有统计回归测试通过 | 否，不应重复实现 |
-| #77-A，heal 选源与输入 | 仍先用 map 首项初始化，再跳过创建默认值；Tag 远端 heal 仍不携带时间戳 | 是 |
-| #77-B，源时间与删除状态 | 六类专用处理函数仍调用到达时间写入路径；部分删除状态不导出 | 是 |
-| #77-D，持续不一致诊断 | 已有损坏配置日志；旧事件静默跳过、默认状态无法选源等诊断未完成 | 是 |
-| Object Lock 错误 wire 字段、接管桶覆盖已有配置 | [#76](https://github.com/pgsty/silo/issues/76)、[#78](https://github.com/pgsty/silo/issues/78) 已关闭，对应 #89、#90 已合并；相关测试本次通过 | 否 |
-| 元数据整记录并发写、删除后重建 | [#103](https://github.com/pgsty/silo/pull/103)、[#156](https://github.com/pgsty/silo/pull/156) 已提供桶锁和保存前物理桶检查；删除后排队写测试本次通过 | 基础已具备，但不能代替同字段时间排序 |
-| 评论提及的 MRF 丢弃观测、delete-marker purge | [#152](https://github.com/pgsty/silo/issues/152)、[#153](https://github.com/pgsty/silo/issues/153) 已于 9 月 9 日随 [#162](https://github.com/pgsty/silo/pull/162) 关闭 | 不再列为 #77 的待实现项 |
-
-#162 明确保留“405 表示 marker 仍存在”的语义，没有采用报告人建议的“405 一律判定永久删除完成”；它也未声称复现外部报告的请求量或持续 405 风暴。本次只核验其合并/关闭状态与提交说明，没有重跑那组故障实验。
-
-最新公开 Release 仍是 `RELEASE.2026-09-03T13-18-01Z`。#77 的状态不能用“已有统计修复”或“已有桶锁”推导为已修复，也不能用较晚 PR 的合并推导为已发布。
-
-## 2. 已复现的真实问题
-
-最直接的例子：源端 10:00 PUT 配置，10:01 DELETE；目标端积压到 10:05 才接收 PUT，并把配置时间写成 10:05。随后接收源时间为 10:01 的 DELETE，判断它比本地旧，返回 HTTP 200，却保留配置。这个例子不需要机器时钟偏差，只需要事件延迟。
-
-根因在 [updateAndParse](https://github.com/pgsty/silo/blob/5c576581631561c446f30ae5b566f0aa793adc1c/cmd/bucket-metadata-sys.go#L129)：它在锁内统一使用 `UTCNow()`；专用 peer handler 在调用它之前，用 getter 返回的时间做比较。
-
-| 配置类型 | peer PUT 保留源时间 | 较新源 DELETE | 本地删除后旧 PUT | 删除时间在元数据导出中可见 |
-| --- | --- | --- | --- | --- |
-| Policy | 否 | HTTP 200，但未删除 | 会复活 | 是 |
-| Tags | 否 | HTTP 200，但未删除 | 会复活 | 否 |
-| SSE | 否 | HTTP 200，但未删除 | 会复活 | 否 |
-| Quota | 否 | HTTP 200，但未删除 | 本次未复活，getter 保留删除时间 | 否 |
-| Versioning | 否 | nil 正确地不修改配置 | 不适用 | 不应套用可删除配置规则 |
-| Object Lock | 否 | nil 正确地不修改配置 | 不适用 | 不应套用可删除配置规则 |
-
-矩阵每项均在 `ErasureSD` 与 16 盘 `Erasure` 两个本地后端执行。六类专用事件通过实际签名 admin 路由调用，检查落盘状态；导出调用实际 `SiteReplicationMetaInfo`。不能把 Quota 与 Policy/Tags/SSE 的 getter 行为写成完全相同。
-
-另外复现了以下路径：
-
-1. **检查与写入不在同一临界区。** 阻塞旧 Tag 事件的锁获取，在锁内提交较新 Tag，再释放旧事件；旧事件仍覆盖新内容。#103 解决跨字段丢更新，没有解决 handler 的 getter → Update 竞态。
-2. **批量元数据入口绕过逐字段排序。** 已持久化较新 Tag 后，发送较旧 bulk 事件，实际 admin 路由返回 HTTP 200，Tag 内容和时间都倒退。该入口被 [桶元数据导入](https://github.com/pgsty/silo/blob/5c576581631561c446f30ae5b566f0aa793adc1c/cmd/admin-bucket-handlers.go#L1106) 使用；[PeerBucketMetadataUpdateHandler](https://github.com/pgsty/silo/blob/5c576581631561c446f30ae5b566f0aa793adc1c/cmd/site-replication.go#L1608) 只检查桶创建时间，非 CORS 字段缺少与已有字段时间的比较。这是旧实施清单漏掉的入口。
-3. **默认值能够成为错误的 heal 来源。** 两站点状态中，一个有真实 Policy，另一个只有更晚的 `CreatedAt == PolicyUpdatedAt` 默认值。最后一次复现分别有 5/32、2/32 轮错误删除有效 Policy，取决于 Go map 遍历顺序。该计数只是复现样本，不能推断生产发生率。
-4. **远端 Tag heal 丢失源时间。** 实际 HTTP 捕获的 `UpdatedAt` 为零，见 [发送字段](https://github.com/pgsty/silo/blob/5c576581631561c446f30ae5b566f0aa793adc1c/cmd/site-replication.go#L5011)。
-5. **同内容、不同时间戳没有同步。** 即便强制 `TagMismatch=true`，内容比较仍使 heal 跳过，较旧排序时间保持不变。两站点看似内容相同，之后却可能对同一个延迟事件做出不同决定。旧方案只统一写入时间、保留全部 heal 跳过条件，仍不完整。
-6. **Quota 的 nil heal 留下缓存。** 向现有 heal 传入显式较新删除状态，磁盘 `QuotaConfigJSON` 被清空，但 `GetQuotaConfig` 仍返回 1024 字节的旧硬配额。这里显式构造了删除状态，因为当前 exporter 正在隐藏它；这是新增 tombstone 导出不能直接交给旧 heal 的具体证据。
-
-从当前 `git blame` 和基线比对看，核心错误继承自 MinIO：heal 初始化逻辑来自 2022 年 `3a64580663`，专用 handler 的旧事件检查来自 2022 年 `7cc9286e0f`，bulk 入口来自 2023 年 `0cde37be50`；0806 基线已包含相关逻辑。此次没有发现其由 SILO 最近的 CORS 或统计修改引入。
-
-对于使用 site replication 的部署，建议按 P1 正确性问题处理：撤销的桶策略可能保留或复活，默认加密和配额也可能与源端不一致。没有启用站点复制的正常单站点请求不触发这些复制路径；本次没有证明对象数据本体丢失。
-
-## 3. 最小且必要的修复范围
-
-**第一部分：收束 A 的选源和输入修复。** 提取一个小的内部选择函数，在初始化候选之前排除零时间和创建默认值；无可信候选时不写入。六类 heal 使用它，补齐 Tag 的源 `UpdatedAt`。Versioning/Object Lock 遇到 nil 保持 no-op，Quota 删除使用正确的删除路径。不能重新做 #76/#78 或把 Lifecycle、CORS 一并重构。
-
-**第二部分：收束 B 的状态更新，包含本次发现的入口遗漏。** 在既有 `metadata.lock` 内完成读取原始字段时间、判旧、写入源时间和保存；普通本地写入保留现有契约。复用 `updateAndParse` 的类型分支及现有 `saveMetadata`，不再从公开 getter 获取删除时间，也不改变 getter 的 S3 错误语义。peer 更新和删除保留各自解析规则，内部返回是否应用，重复且相同的状态不重复保存。
-
-本地 heal 必须使用同一排序路径，不能选好源后调用普通 `Update/Delete` 再生成到达时间。批量复制入口也必须在其现有整记录锁内逐字段判旧；bulk 的 nil 继续表示“未提供此字段”，不能改成批量删除。heal 需要同步较新的时间戳，即便内容已经一致；这不要求重写 #91 的计数或增加公开 API。
-
-当前 `saveMetadata` 已在锁内检查物理桶存在，桶删除也使用同一把锁。8 月方案里另加一套 `peer-require-existing` 防幽灵桶机制已没有必要，应保留并复用现有保护。确有 `Created == 0` 的历史桶如何补齐创建时间仍应覆盖，但不能退化当前所有写入都执行的存在性检查。
-
-**第三部分：单独启用删除时间导出，补足轻量诊断。** Policy 已导出时间；实际新增的是 Tags/SSE/Quota 的 nil payload 时间。先使接收与 heal 能正确处理删除，再开放这些状态。混合旧版 SILO 的风险已经有 Quota 缓存复现支撑；这是产品自身的滚动升级问题，不是要求兼容未经修改的上游 MinIO。
-
-可以使用仅针对这项导出的明确 opt-in：默认关闭新增导出，全站点升级到具备 A/B 修复的版本后启用。若项目选择自动能力协商，可后续单独实现；不必为了它阻塞没有新增 wire 字段的选源与接收端修复，也不需要为本 Issue 建一个通用能力框架。具体开关尚未实现，不应把这里的建议当成现有配置。
-
-诊断只需解释“过旧事件”“早于桶创建”“只有默认/未知候选”等跳过原因，使用现有有界去重/限频设施，保持 RPC 的既有成功语义。不需要另建重试队列、每轮日志或大型监控系统。
-
-有三项边界必须在实现中明确：
-
- 旧版 Tag heal 会发送零时间。新实现不能无说明地全部拒绝；可保留明确的兼容降级路径，在旧节点存在时不承诺完整排序收敛。
- 相同时间戳、不同内容的冲突：只给 heal 增加 deployment ID 平局规则，并不能让到达顺序不同的 peer apply 本身确定。若承诺此类冲突也收敛，apply/heal 应共用一个类型内比较规则；可参考 CORS 的删除优先和载荷排序，不需要新增持久化源站 ID。若不纳入本轮，必须列为剩余边界，不能称全量收敛审计完成。
- 旧版本已经写坏的到达时间无法从现有记录还原。升级不会自动恢复历史操作顺序；应由操作者确认权威状态，并在升级和时钟检查后重新提交相关配置/删除，再验证各站点。
-
-因此，建议交付为有限的选源修复、统一源时间更新、删除导出与诊断三个可审核部分。已有锁、物理桶保护和 C 的计数修复直接复用。预期产品修改集中在 Server 的 site-replication 与 bucket-metadata 路径，无须为核心修复升级 Console/mcli/silo-pkg、改存储格式或重写复制架构。
-
-## 4. 验证证据与关闭条件
-
-本次最后一轮执行：
-
-```text
-GOWORK=off go test -tags kqueue,dev ./cmd \
-  -run 'TestIssue77Current|TestSiteReplicationStatusAccountsPerSiteAndSurvivesMalformedConfig|TestPeerBucketObjectLockMetadata|TestPeerBucketAdoption|TestQueuedMetadataUpdateAfterDelete' \
-  -count=1 -timeout=5m -v
-```
-
-四组审计复现测试包含 24 个后端/场景子测试，均暴露预期的现存缺陷；八个现有回归测试通过。总命令退出码为 1，原因是上述针对期望正确行为的断言失败，不代表修复验收通过。未运行全量测试、race suite、真实多站点中断/重启或滚动升级实验。
-
-保留的证据：
-
- [完整运行日志](issue-77/current-tests.log)
- [源时间、删除和 bulk 审计测试](issue-77/issue77_review_test.go.txt)
- [heal 与锁竞态审计测试](issue-77/issue77_heal_review_test.go.txt)
-
-测试以 `.go.txt` 保存，避免将故意失败的审计用例加入正常 Go 测试集。可在上述 SHA 的独立 worktree 中复制为 `cmd/issue77_review_test.go` 和 `cmd/issue77_heal_review_test.go` 后重跑。创建默认值测试使用重复 map 遍历来观测缺陷，正式回归应在抽出选择函数后改成确定性用例。
-
-关闭 #77 前需要：上述失败场景转为正确行为；分别覆盖六类配置、落盘重载、漏发/重复/乱序、同字段竞态、同内容新时间、批量导入入口、legacy 桶、删除后的桶及混合版本。全站点升级后的实际断线重连/重启验证应保留独立证据。当前结论是“缺陷与实施范围已核实”，不是“修复完成”。
@@ -1,123 +0,0 @@
-# Issue #77 最小充分修复计划
-
-> 历史快照（2026-09-12，实施前）。其中的 Issue 状态、待办和方案约束只描述当时情况；最终实现、修正与验收以 [归档总记录](issue-77.md) 为准。
-
-版本：v4 定稿，2026-09-12。状态：作者复核完成，Claude Code `claude-opus-5 --effort max` 四轮实际评审后，最终结论为 **GO_WITH_NONBLOCKING_NOTES，实施前阻断 0**。非阻断说明已落实到本计划，见 [最终评审汇总](issue-77/review/final-review.md)。
-
-基准：`pgsty/silo` main `5c576581631561c446f30ae5b566f0aa793adc1c`。本轮交付是方案与评审，不表示已经实现、合并或发布。问题证据见 [当前核验](issue-77-current.md)，意见处置见 [首轮处置](issue-77/review/decisions-v2.md)、[第二轮处置](issue-77/review/decisions-v3.md)、[接管边界补查](issue-77/review/decisions-v4.md)。
-
-## 目标与边界
-
-修复 Policy、Tags、SSE、Quota、Versioning、Object Lock 六类桶配置的源时间丢失、锁外判旧、heal 错选源和删除传播不完整。覆盖普通本地写入、专用 peer 事件、bulk/import、local/remote heal、元数据导出与初次同步。
-
-在修复 PGSTY 栈、同一已知桶世代、合法带源时间事件的范围内，使重复、乱序和漏发后的状态能够确定性收敛。开关关闭期间新增删除信息不可见，不承诺完整删除收敛；无时间旧事件、旧版污染时间和桶创建世代冲突需要单独解释，不能自动推断历史真相。
-
-不重做 #91 的计数、#76 的 Object Lock wire 修复、#78 的桶接管、#103/#156 的锁与删除保护。不改变 CORS、Lifecycle/expiry、notification、对象复制、MRF、resync、IAM 的语义；不改存储 schema、SDK、Console、mcli 或 silo-pkg，不新建能力协商、复制框架、锁或重试系统。
-
-## 行为契约
-
-### 1. 明确事件、缺省值与删除
-
-| 类型 | 专用事件 | bulk 字段未提供 | bulk 字段明确提供 |
-| --- | --- | --- | --- |
-| Policy | nil 为删除；沿用现有解析器 `IsEmpty()` 为删除 | 保留 | 非空 RawMessage 按现有解析器处理；语义空策略归一为删除 |
-| Quota | nil 为删除；非 nil 按现有 quota 解析器处理 | 保留 | 非空 RawMessage 按现有解析器处理，零 quota 仍是 live 文档 |
-| Tags / SSE | nil 或 base64 解码后空内容为删除 | 保留 | 空字符串为删除，其他内容按原规则解码、校验 |
-| Versioning / Object Lock | nil/空内容为 no-op | 保留 | nil/空内容仍为 no-op，不能清空配置 |
-
-“未提供”必须依据真实 wire 类型判定：Policy/Quota 的 `json.RawMessage` 为 nil 或空切片时经 `omitempty` 省略；**显式 JSON `null` 解码后是非 nil 的 `[]byte("null")`**，不能与缺省混淆。按现有解析器，Policy `null` 是语义空策略，Quota `null` 是零值 quota 文档。`*string` 类型的 JSON `null` 则解码为 nil。不得使用统一的 `len(payload)==0 => 删除` 来处理 bulk。
-
-合法 `{}`/零值 quota 保持 live；取消普通 quota PUT 出站时“零配额改写为 nil”的逻辑，保存与发送同一个语义状态。Policy 保留现有专用 peer 的“空策略=删除”解释，本地 PUT、导入和 bulk 统一归一为同样的删除状态；这是需要写入兼容说明的小范围变化：空策略本地立即按删除处理，GET 返回既有 NotFound 行为，不再先保留空文档、等复制后才被清除。
-
-**同一次操作的落盘状态与出站事件，经同一归一规则后，必须具有相同的 `(kind, payload key, source time)`。** JSON 的无意义编码次序不要求字节相同；Object Lock 改写后的有效 Versioning 文档必须来自提交结果。
-
-### 2. 统一状态与排序
-
-内部仅需一个小状态表示：baseline / live / tombstone，以及比较键和字段源时间。baseline 可携带历史有效配置，但没有真实修改时间；它绝不能被当成删除。复用 CORS 已有设计思路，不改 CORS 本身或扩展为通用框架。
-
- 使用 `BucketMetadata` 原始字段时间；字段时间为零时在比较视图中补为 Created，与 `defaultTimestamps()` 一致，不借用会隐藏墓碑时间的 getter 或整记录 `lastUpdate()`。
- 在已知 Created 下，零字段时间回退后等于 Created 的状态是 baseline：有有效非空配置的 baseline-live 可作为初始化候选；空 baseline 只是缺省值，不能作为删除或 heal 来源。真正严格早于自身 Created 的字段不是候选；Versioning/Object Lock 的空候选无论时间如何都不参与选源。真实 live/tombstone 的时间必须大于 Created。
- 专用 peer 的零时间保留兼容例外：锁内按本地新操作分配时间并限频记录 legacy-zero；不受删除传播开关影响，不在源时间排序保证之内。bulk 零时间仍按现状拒绝。
- 非零事件**严格早于**目标桶 Created 才成功/no-op 并记录 before-created，heal 对这种目标跳过。等于目标 Created 的 live 事件可更新仍是 baseline 的目标字段，时间仍保存为源 Created，不盖上到达时间；同时间的空/nil 只算空 baseline，不能删除配置。不同桶世代不能自动合并，需要运维处理，不纳入收敛承诺。
- 排序先比较是否为真实状态：任何真实 live/tombstone 都胜 baseline，不能让较晚创建的默认值压过较早的真实修改。真实状态之间再比较源时间，同时间 tombstone 胜 live、live/live 的**稳定比较键字节序较大者胜**。baseline-live 只胜空 baseline，或在 baseline-live 之间按同一稳定键较大者胜；永远不能覆盖真实 live/tombstone。所有候选都是 baseline-live 时仍可确定性初始化并收敛；全部为空 baseline 才安静 found=false。键、状态级别和源时间相同为 no-op。deployment ID 不参与上述比较，也不作为新字段保存。
- Quota 使用现有解析结果的 JSON 编码作为比较键。Policy 在 Server 内生成确定性比较表示：既有解析器校验/去重后，对其**完整 JSON 树**的对象键和集合数组递归排序，统一覆盖 Statement、Action/NotAction、Resource/NotResource、Principal、Condition；保留数字类型和精度。比较键必须是已解析策略的纯函数，同一配置从两个站点分别解析也必须得到相同键。普通 `json.Marshal(BucketPolicy)` 不稳定，不能直接作键；不增加依赖不支持的 NotPrincipal 语法，不改 wire/schema。新状态比较仅使用这一套键：不把忽略 Sid 且对 Statement 顺序敏感的 `BucketPolicy.Equals` 再叠加为另一套判等规则，既有公开统计对 Equals 的使用保持不变。
- XML 使用有效文档字节，保留大小写和实际内容；Versioning 先应用下述现有 Object Lock 约束。比较器不能依靠字节序方向来补偿保存阶段的隐式改写。
-
-### 3. 先得到有效状态，再比较与提交
-
-原始读取、类型处理、比较、修改及保存均在现有 `metadata.lock` 内。内部入口必须传递现有 lock context，避免 legacy migration 再次取锁。
-
- 提取并复用 `parseAllConfigs` 已有的 Object Lock→Enabled Versioning 归一规则，使比较视图和真正保存一致；不得扩大 suspend、prefix exclusion、retention 的限制。bulk 先确定实际接受的 Object Lock，再比较该约束下的 Versioning，并在最终提交前应用同一规则。
- 以归一化后的有效 `(kind, key, time)` 判定变化；更新了时间也算变化。完全重复不保存、不通知；一次 bulk 校验失败不保存部分结果；成功至多保存一次，解锁后通知。
- 可删除字段清空必须基于现有 parse=false 的新加载对象，不能在已经解析且仍持有旧 quota 的对象上执行 Update(nil)。bulk 保留原始读取再解析保存的方式。不要顺手修改 Quota getter 或所有 `parseAllConfigs` 空分支。
- 保存函数必须让需要发送 hook 的调用方拿到**本次提交的最终快照**，不能先解锁再读取“最新”状态拼接旧时间。最小做法是让内部 `saveMetadata` 接收元数据指针并回写 Save 的归一化结果，机械更新现有少量调用；对外 `Update/Delete` 签名不变，新增内部提交结果仅供需要该快照的本地 handler/import 使用。不得原地修改已发布到缓存的引用字段。
- 保留现有物理桶存在检查、删除锁序、迁移、后台通知上下文。真实历史桶 Created 为零时仅走少见的物理桶 Created 补齐路径；物理桶缺失返回现有错误；创建时间仍未知则不伪造到达时间，报告 indeterminate。
-
-## 三个实现提交
-
-### 提交 1：原子 apply、发送一致性与本地时间
-
-主要文件：`cmd/bucket-metadata-sys.go`、`cmd/bucket-metadata.go`、`cmd/site-replication.go`、`cmd/admin-bucket-handlers.go`，以及实际需要提交快照的本地配置 handler。
-
-1. 在现有 update/delete 内部路径增加源时间、状态比较和提交结果；公开签名不变。六个 peer handler 移除锁外 getter 判旧，锁内持久化原始源时间。保留 Object Lock 的 legacy Tags 字段载荷回退。
-2. bulk 对明确提供的六类字段在已有锁内逐字段比较，再原子保存；未提供字段不动，不能循环调用会重入锁的公开 handler。保留 CORS 独立分支与既有行为。
-3. 六类本地实际写入在锁内分配 `max(UTCNow(), Created+1ns, 当前字段时间+1ns)`。其他类型不变。`enablePeerBucketVersioning` 的实际变更也使用它，只有缺失配置的创建 bootstrap 继续 Created 默认值。
-   接管已有桶时保留原 Created，再执行现有 `SetCreatedAt`；如果 Created 改变，仅将这六类中原本为零或等于原 Created 的默认字段时间调整到新 Created，随后再做既有 versioning/lock bootstrap。原本晚于旧 Created 的真实修改/删除时间及其配置保持不变。不能仅凭 payload 为 nil 判断默认值。该小分支防止 Created 前移时默认值变成假墓碑、后移时历史初值被误判为无效，复用已有接管锁，不重做 #78 的配置保护。代码注释明确限定六类的原因；本轮不改新比较器未读取的其它配置时间。
-4. quota 本地 PUT 保留零值文档并原样表示该语义；Policy 空策略本地与 peer 一致走删除。需要归一化的本地 handler 从本次提交快照生成 hook；其他内容不发生归一变化的路径可保留既有编码，但必须满足三元组一致性。
-5. 导入在每桶最终提交锁内，为本次涉及的六类字段生成共同 commitAt，严格大于 Created 和这些字段当前时间且不早于锁内现在。该时间同时用于落盘和 bulk hook，不能沿用 ZIP 开始时间。bulk hook 从最终提交快照构建；若导入的空 Policy 已归一成删除，另外使用现有专用 Policy nil 事件表达它，不能因 `omitempty` 漏发。未导入字段不改，Object Lock 的既有派生 Versioning 修正保留原时间语义；CORS 继续独立时间/事件，其他字段不参与该上界。
-
-完成条件：六类源时间落盘；旧事件不能越过锁覆盖新状态；四类删除不会被旧 PUT 复活；真实 wire、落盘和出站状态一致；重复无写入；bulk 与 import 没有绕过排序或静默遗漏删除。
-
-### 提交 2：heal 选源与应用同规则
-
-主要文件：`cmd/site-replication.go`。
-
-1. 先过滤空 baseline、无效来源、严格早于自身 Created 的字段和 update-only 空配置，再按上述强弱排序选最大状态；无候选必须显式返回 found=false。保留历史 baseline-live 的初次同步和 heal 能力，消除六处“先 seed map 首项，再过滤默认值”的写法。
-2. 选源和目标遍历都跳过 `info.Sites` 中不存在的 deployment ID，包括不可达站点的空 ID 占位项；单一 peer 失败记录后继续其它目标，不因 map 顺序放弃健康站点。不改变状态计数或新建重试机制。
-3. 本地 heal 使用提交 1 的源时间 update/delete；远端仍用原有逐类型 RPC，全部携带源时间，补齐 Tag 的 UpdatedAt。
-4. 比较完整有效状态，去掉公开 mismatch/payload-only 对写入的门控；同内容较旧时间也同步。对已归一且相同的目标不写入、不发 RPC。Versioning 比较使用与该站点 Object Lock 一致的有效文档；全站点 Lock 状态补齐后不再因旧原始文档产生空转。
-5. 保留 #91 的计数与公开字段；创建世代冲突、无可用来源通过有限诊断解释。已知 baseline 且各站点无实质差异时安静 no-op。
-
-完成条件：map 顺序不影响结果；默认空值不再删真配置；历史 baseline-live 可以初始化新站点、不能覆盖真实状态；同内容不同时间、同时间冲突最终一致；Tag 时间完整；Quota 删除后磁盘、缓存、重载一致；不可达占位项不阻断健康目标；完整状态已可见且稳定时，第二轮 heal 无写入/广播。
-
-### 提交 3：新增删除传播与有界诊断
-
-只增加一个启动开关，暂定 `MINIO_SITE_REPLICATION_METADATA_TOMBSTONES=off/on`，默认 off；实现沿用现有 env 开关写法，不做能力协商。
-
-| 行为 | off：升级阶段 | on：所有参与节点修复后 |
-| --- | --- | --- |
-| 带时间 peer apply、锁内排序与 heal | 使用提交 1/2 | 同左 |
-| 专用事件零时间 | 兼容应用并记录 legacy-zero | 同左，不新增协议拒绝 |
-| Tags/SSE/Quota nil payload 时间导出 | 保留旧版条件导出 | 导出 `time > Created` 的真实删除时间 |
-| Policy 时间导出 | 保留已有行为 | 保留已有行为 |
-| 初次同步的真实删除 | 保留已有行为 | Policy/Tags/SSE/Quota 都发送专用 nil + source time 事件 |
-
-开关只控制**新增**删除信息的导出/初次发送，普通本地删除事件照常复制。**off 不等于禁止删除传播：Policy 墓碑原本已导出，修复后在 off 下也照常参与 heal；Tags/SSE/Quota 的新增墓碑信息才被门控。** off 期间这些字段的隐藏墓碑会使 heal 继续尝试过时 RPC，由接收端排序拒绝；这是状态不可见时的已知代价，不承诺第二轮零 RPC，也不为这种正常拒绝增加每轮日志。
-
-开关不检测或证明远端能力。启用条件是所有参与站点的全部节点已经修复，同一站点配置一致，旧请求排空；旧节点仍在线时保持 off。此隔离有实证依据：旧版接到新增 Quota heal 墓碑会留下已解析缓存残留。
-
-日志仅保留三个实际原因：legacy-zero、before-created、indeterminate（未知创建时间、缺失/不可达来源或有实际差异却无可用候选）。精确重复、正常旧事件和成功裁决的同时间冲突不记警告。复用 `LogOnceIf`，以稳定的桶/字段/原因作为 key，**错误正文也必须稳定**；变化的时间与 peer 详情放入日志 ReqInfo，沿用现有每小时清理，不新增限流框架、不输出完整策略。
-
-Server 文档解释启用顺序和回滚：降级前所有修复节点先关开关，然后滚动降级；旧软件缺陷会恢复。点名旧版 Tag heal 无 UpdatedAt 的来源。旧版到达时间污染、legacy-zero 产生的新本地时间以及创建世代分歧无法自动还原；操作者查看状态后在权威站点重新提交需要纠正的配置/删除。历史世代冲突先处理桶身份，不能靠任意站点强刷绕过创建保护。
-
-## 最小验收矩阵
-
-| 组 | 必须覆盖 | 证据方式 |
-| --- | --- | --- |
-| T1 | 六类 PUT 源时间；四类 DELETE；旧事件不回退；重复无写入；零 quota 三元组一致；带合法 Version 的空 Policy PUT 成功、GET NotFound、专用删除事件与落盘一致 | 真实 admin/S3 路由、ErasureSD/Erasure16、磁盘重载、RPC 捕获 |
-| T2 | 同时间两种到达顺序结果相同、删除优先；Policy 多集合及 NotAction/NotResource/Condition 反复编码与排列后，两站点独立解析得到相同键；Sid 差异及大整数不被错误合并 | 确定性比较器测试与代表性真实 handler |
-| T3 | Versioning/Object Lock nil/空 no-op；legacy Tags 载荷回退；Object Lock + prefix exclusion/ExcludeFolders 在普通写、peer、bulk/import 后有效状态一致，第二轮 heal 无额外写入 | 原有 #76/#78 回归加针对性用例 |
-| T4 | 锁前旧事件排队、较新写先提交后旧事件不得覆盖；不同字段并发均保留 | 复用已有 `lockBucketMetadataAcquireHook` / RMW 屏障、两种 ObjectLayer、目标 race |
-| T5 | bulk 新/旧/缺省字段混合；真实 JSON 编解码的 nil、空 RawMessage、显式 null、空字符串、空策略、零 quota；非法字段不部分保存；只导入 tags 不修改 Policy/Quota | 真实 bulk 路由、缓存与磁盘 |
-| T6 | 本地时间胜过已有未来时间；相邻提交不倒序；ZIP 导入期间插入写入，最终落盘和发出事件的状态与时间一致，空 Policy 删除不漏发 | 本地 API、import 路由与 RPC 捕获 |
-| T7 | map 排列、空 baseline/全无候选；baseline-live 初始化与同级冲突收敛，但不能覆盖真实 live/tombstone；nil@Created 不删除配置；空 update-only；Tag 时间、Quota 清缓存、空 ID、世代冲突 | 确定性 heal 本地/远端用例 |
-| T8 | 墓碑经保存/缓存失效/重启仍有效；缺桶、排队写入、零 Created；历史字段时间等于 Created 的桶经初次同步、一轮 heal 后一致，第二轮无写入/广播；接管 Created 前移/后移/不变时默认时间仍是 baseline，真实 PUT/DELETE 时间不变；on/off 与 legacy-zero，off 下 Policy 仍 heal、其它隐藏墓碑允许被拒 RPC 但无每轮警告 | ObjectLayer 与进程内旧版 wire/SRInfo 模拟；复用既有删除/迁移回归 |
-| T9 | 修复版双站点短暂断线、漏发/重复/乱序后六类合法配置收敛；删除传播启用后第二轮稳定无写入；日志确实有界 | 隔离双站点实验，不把未知桶世代或零时间事件算成通过 |
-
-固定旧版与修复版混合进程仅作一次性升级冒烟，不新增为长期提交门槛。将已有审计用例改成正式确定性回归，不能把遍历 map 的概率复现直接提交。先记录未修复失败，再验证通过；复用原有 Object Lock、adoption、metadata-lock、计数、CORS 回归，并在实现后运行目标 race、仓库必需检查和完整 Go CI。方案审查、局部测试、双站点结果、合并与发布是不同证据。
-
-## 作者复核结论
-
-三个提交分别处理写入正确性、heal 收敛和新增删除信息的升级边界；每项对应已证实缺陷或本次修复直接触及的实际路径。缩减了拒绝零时间的新协议行为、混合版本长期测试门槛和新锁屏障；保留已有持久化、通知与缓存语义。
-
-Opus 第二轮已确认首轮 R1/R2/R3 实质关闭，第三轮确认历史桶初始化边界，第四轮确认接管默认时间修正必要、充分且未扩大范围；最后两轮均为 0 阻断。四轮原文、版本快照、逐项处置与模型调用证据均保留。方案可以进入实现；当前只完成方案和诊断，仍须通过上述实现期验证，不能据此认定问题已修复或可关闭。
@@ -1,85 +0,0 @@
-# #77：桶配置复制修复与验收归档
-
-归档日期：2026-09-12。对象为 [SILO #77](https://github.com/pgsty/silo/issues/77)。问题真实；来源时间、删除状态和 heal 选源共同决定是否收敛，不能只补一个删除分支。最终实际 Claude Code Opus 5 Max 实现审查结论为 **GO_WITH_NONBLOCKING_NOTES，阻断 0**，完整 cmd 和最终 lint 随后通过。
-
-本文固定研究与验收时的事实。合并状态以 Issue 关联 PR 为准；主干包含代码不等于镜像、软件包或生产部署已经更新。研究叙述见伴生站的[中文设计记录](https://github.com/pgsty/silo.pgsty.com/blob/main/content/blog/design/bucket-metadata-convergence.zh.md)和[英文设计记录](https://github.com/pgsty/silo.pgsty.com/blob/main/content/blog/design/bucket-metadata-convergence.md)，操作契约见 [Server site-replication README](../site-replication/README.md)。
-
-## 实现为什么最小、必要且足够
-
-源端 10:00 PUT、10:01 DELETE，目标到 10:05 才收到 PUT。旧实现把字段时间写成 10:05，随后把源时间为 10:01 的删除当作旧事件跳过，返回成功但保留配置。删除后没有导出时间的字段，在漏发后还会被对端旧值恢复。另有锁外判旧、批量入口绕过排序、默认值抢占来源、Tags heal 丢失时间和同内容不更新时间等独立入口。
-
-| 必要改动 | 少了它会发生什么 | 复用的边界 |
-| --- | --- | --- |
-| 在现有整桶锁内读取、比较并保存来源状态 | 锁外判旧仍可覆盖并发的新状态；到达时间继续污染排序 | 既有 `.metadata.bin`、`metadata.lock` 和物理桶存在检查 |
-| Policy、Tags、SSE、Quota、Versioning、Object Lock 共用确定性比较 | 等时冲突仍依赖到达顺序；heal 与接收端可能选出不同结果 | 既有载荷、字段时间及 Created，不增加 wire 或持久化字段 |
-| 专用事件、bulk、导入、本地写和 heal 都使用提交后的状态 | 只改一个 handler 会留下旁路；归一化后的 Versioning/Quota 与出站事件可能不同 | 既有解析、保存与复制钩子，bulk 仍是一次原子保存 |
-| 删除导出开关默认关闭 | 直接对旧版节点导出新增删除状态有已复现的 Quota 缓存风险 | 全部参与节点升级后统一启用，不引入能力协商框架 |
-| 物理 Created 恢复与初次同步传播 | 历史无创建时间的真实桶会失去六类配置写入能力 | 现有物理探测；目录 mtime 只是近似值，不推断真实桶世代 |
-| Policy 状态键与 heal 一致；异常按原因去重 | 永久假 mismatch 无法修复；正常空基线噪声或无来源异常静默 | 既有每站点统计与 logger，不增加后台协调系统 |
-
-真实状态优先于创建基线；真实状态按来源时间、等时删除优先、规范内容键排序。历史 `live@Created` 可以初始化空目标，但空基线不能删除真实配置。本地纠正时间在锁内取 `max(now, Created+1ns, fieldTime+1ns)`；带时间 peer 事件保留来源时间。Versioning/Object Lock 的 nil 保持 no-op，Quota `null`/`{}` 保持零配额配置，空 Policy 归一为删除。
-
-Policy 编码器保留在 GET/export/peer 是必要的：既有解析器可以保存的负集合策略必须能读回和复制。PUT/import 继续使用同一编码器，避免额外表示分支。没有为了本轮重写 CORS、Lifecycle、对象复制、IAM 或 #91 已完成的计数，也没有修改 SDK、模块依赖或升级协议。
-
-充分性限定在同一桶世代、有效且可排序的来源状态、全部参与节点升级并开启删除导出的范围内。历史时间污染、零时间兼容事件和桶身份冲突不属于自动恢复承诺。
-
-## 计划与实际对抗审查
-
- [实施前核验与失败复现](issue-77-current.md)、[最终 v4 计划](issue-77-plan.md)。这些文件保留历史状态，不能当成当前待办。
- [四轮计划审查与意见处置](issue-77/review/final-review.md)：前两轮 NO_GO，后两轮零阻断；保留各版计划、完整最终意见、调用元数据和必要探针。
- 三轮实现审查均使用实际 Claude Code `claude-opus-5 --effort max`。模型身份取自实际 assistant 消息，effort 取自显式调用参数。源码在独立快照中审查，审查者读代码和执行结果，没有代为运行这些验收。
-
-| 实现审查 | 固定源码 | 结论与处置 |
-| --- | --- | --- |
-| [首轮完整意见](issue-77/implementation-review/opus5-max-review.md) / [调用记录](issue-77/implementation-review/session.json) | `4089113e3` | GO_WITH_NONBLOCKING_NOTES，条件性阻断 F1；物理 Created、Policy 假 mismatch、诊断和证据缺口随后修复 |
-| [第二轮意见](issue-77/implementation-review/round-2/opus5-max-review.md) / [调用记录](issue-77/implementation-review/round-2/session.json) | `62cf066ff` | 条件性和无条件阻断均为 0；修正首轮对 Policy 编码器的过宽质疑，进一步补齐无来源诊断和初次同步回归 |
-| [最终定向意见](issue-77/implementation-review/round-3/opus5-max-review.md) / [调用记录](issue-77/implementation-review/round-3/session.json) | `fcbb93e89` | 阻断 0，确认后续 `461e9a721` 的测试改写等价；读取时尚在运行的 cmd/lint 后续均退出 0 |
-
-最初实现审查 SHA `4089113e3` 补签 DCO 后对应 `1ee64a8d8`，两者树完全相同。生产代码最终固定在 `fcbb93e8957275bfa7ad4e6154e93f4d7b0a7025`；验收源码 HEAD 为 `461e9a721047c63e1a95f54ad4b533a6b89def30`，只在两个测试文件有格式和等价条件改写，见 [tree 对照](issue-77/implementation-review/round-3/final-tree-equivalence.json)与 [patch](issue-77/implementation-review/round-3/test-style.diff)。本次归档不改变生产代码或正式回归测试。
-
-审查原文保留当时的判断，不将后续作者修正倒写成评审者已经观察到的结果。首轮误读 README 中既有的空 Policy / 零 Quota 说明，第二轮修正编码器必要性的判断；最终处置以本文和伴生站的逐项表为准。
-
-## 验收证据
-
-| 检查 | 执行对象与结果 |
-| --- | --- |
-| [完整 cmd](issue-77/implementation-review/round-3/final-cmd.log) / [命令及退出码](issue-77/implementation-review/round-3/final-cmd.json) | `fcbb93e89`，CGO=0，全部通过，包测试 492.776 秒 |
-| [最终目标 race](issue-77/implementation-review/round-3/final-target-race.log) / [命令](issue-77/implementation-review/round-3/final-target-race.json) | `fcbb93e89`，真实创建时间、诊断、初次同步、接管、Policy 状态和全部 CORS 命名用例通过 |
-| [测试改写后的 race](issue-77/implementation-review/round-3/final-style-target-race.log) | `461e9a721`，受影响用例通过 |
-| [build](issue-77/implementation-review/round-3/final-build.json)、[vet](issue-77/implementation-review/round-3/final-vet.json)、[lint](issue-77/implementation-review/round-3/final-lint.json) | 最终生产树 build/vet 通过，`461e9a721` lint 零问题；可选 typos 未安装，按 Makefile 跳过 |
-| internal、S3 Select race、生成文件、兼容检查、六平台编译 | `62cf066ff` 的 `ci-*-after.log` 作为较早阶段的补充证据，不冒充最终树在全部平台运行通过；归档中的空生成日志本身不证明退出状态 |
-| [两个真实站点进程](issue-77/implementation-review/round-3/final-runtime.log) / [执行元数据](issue-77/implementation-review/round-3/final-runtime.json) | 干净 `fcbb93e89` 构建：六类历史配置、真实漏发恢复、乱序、四类删除跨重启、两次各 65 秒零 metadata RPC、异常去重、gate=off 与固定旧版的 PUT/DELETE 冒烟均通过 |
-| [伴生站构建](issue-77/implementation-review/round-3/docs-check.log) | 提交为 `9fa6248` 的文章内容通过 Hugo 严格构建和站内链接检查，EN 1207 / ZH 1219 页；后续合并状态文案单独检查 |
-
-[归档清单](issue-77/archive-manifest.json)保存每份材料的原始和归档后 SHA-256，以及全部改动源码的 SHA-256。[二进制身份](issue-77/implementation-review/round-3/binary-identity.json)保留实际版本、Go 构建身份和摘要：最终运行二进制来自干净 `fcbb93e89`，SHA-256 为 `4825a801ce0ac48d636d9428ce4cd5c17a20b6cfec0b569de70a124c9c005049`；旧版来自干净基线 `5c5765816`。第三轮审查的 cmd/lint 条件由上述已完成的 JSON 关闭。
-
-### 反向复现与排除的证据
-
- [confirmed-before.log](issue-77/implementation-review/round-2/confirmed-before.log)：正式测试叠加 `1ee64a8d8` 的 `erasure-server-pool.go` / `site-replication.go`，真实 ObjectLayer 创建时间恢复及旧 Policy 顺序失败；[修复后](issue-77/implementation-review/round-2/confirmed-after.log)和 [race](issue-77/implementation-review/round-2/confirmed-after-race.log)通过。
- [confirmed-diagnostics-before.log](issue-77/implementation-review/round-2/confirmed-diagnostics-before.log)：叠加旧诊断路径，普通空基线误报。第一处断言已经终止测试，不能声称它同时证明后面的 Warning 级别和去重断言在旧码失败。
- [no-source-before.log](issue-77/implementation-review/round-2/no-source-before.log)：直接运行 `62cf066ff`，三种已有无效状态在没有有效来源时缺少诊断；这一次不是旧码 overlay。
- [initial-sync-before.log](issue-77/implementation-review/round-2/initial-sync-before.log)：保留物理恢复修复，仅叠加 `1ee64a8d8` 的 `site-replication.go`，首次同步出站仍传零 Created。该单测的 peer 只确认 RPC，不证明真实对端或本地 peer 分支已经落盘。
-
-`findings-before.log` / `findings-after-1.log` 是早期夹具失败，不能证明产品缺陷；过渡阶段的旧 helper 遗留引用和测试格式失败也不能当成通过记录。这些草稿不进入本归档的验收证据。另一次 Codex 调用因使用限额失败，不计入完成审查次数。
-
-### 重跑真实双站点验收
-
-[独立 Go 驱动](issue-77/runtime/main.go)与原执行内容相同，仅添加 `ignore` 构建标记及说明，避免进入普通包测试。它只启动回环地址上的一次性实验实例，每站四个数据目录，使用文件内声明的专用实验凭据。给它一个全新目录和自行从固定提交构建的两个二进制：
-
-```sh
-go run docs/investigations/issue-77/runtime/main.go \
-  /absolute/new-lab-directory /absolute/silo-fixed /absolute/silo-before
-```
-
-固定构建分别为 `fcbb93e89` 和 `5c5765816`，可在独立干净 worktree 使用 `CGO_ENABLED=0 go build -trimpath`；记录 `--version`、`go version -m` 和 SHA-256。省略旧版二进制参数会跳过混合版本冒烟，不能将其报告为执行过。最终删除收敛快照见 [converged.json](issue-77/runtime/converged.json)，数组顺序由驱动的 `states` 函数定义。
-
-原执行环境为 `go1.27.1 darwin/arm64`。因宿主盘可用空间比例触发存储保留阈值，完整 cmd 和进程实验使用独立 16 GiB APFS 测试卷，没有降低生产阈值。临时卷、审查 worktree 和实例进程均已清理。上述是本地可复核的执行证据，不替代 GitHub Actions、Linux 多节点集群、发布制品或生产验证。
-
-## 剩余边界
-
-1. `MINIO_SITE_REPLICATION_METADATA_TOMBSTONES=off` 是默认值。全部站点全部节点升级、配置一致并排空旧请求后，统一开启并重启，才具有新增 Tags/SSE/Quota 删除的漏发自愈能力。普通删除事件与原有 Policy 删除导出仍保留。
-2. 到达时间污染、零时间旧事件和桶世代冲突无法凭现有数据重建。物理 Created 只是目录 mtime 的近似值；较早事件仍跳过，由操作者核对权威状态后重新提交纠正。
-3. 日志按每桶/字段/原因去重，仍可能随桶和有值字段数量增长。本轮不增加全站日志调度。
-4. 最终审查保留三项非阻断改进：解码失败诊断中的时间可能为零；初次同步单测没有证明本地 peer 分支落盘；未来并发后台日志可能需要更强的测试隔离。生产路径已复核，本轮不为这些建议新增 helper 或 hook。
-
-归档保留最终报告与调用身份，不发布原始模型推理流、二进制、实验数据卷或无效夹具日志。工作站路径、历史文档链接与非必要 JSON 字段经过整理；哈希清单区分原始产物与归档副本，不能将整理后的报告哈希冒充原文件哈希。
@@ -1,437 +0,0 @@
-{
-  "archived_at": "2026-09-12",
-  "issue": "https://github.com/pgsty/silo/issues/77",
-  "base": "5c576581631561c446f30ae5b566f0aa793adc1c",
-  "tested_source_head": "461e9a721047c63e1a95f54ad4b533a6b89def30",
-  "reviewed_production_head": "fcbb93e8957275bfa7ad4e6154e93f4d7b0a7025",
-  "source_sha256": {
-    "buildscripts/rebrand-guard/compat-baseline.json": "7ea6485008dd998373144ebded82483fe2c33ae60712abcf2fba973a7506b3e1",
-    "cmd/admin-bucket-handlers.go": "3aa8fc80cb8f2c4725acb8fbec026d47cd71a60ad58d06f905aa013540b30db7",
-    "cmd/bucket-cors-site-replication_test.go": "2cff52d3857e912526aaae40736b077e6b8bd640e098934b8af0ec76354f3747",
-    "cmd/bucket-metadata-replication.go": "07730f6a19b78a9b51dd9d56696812299b3aedce1a779b23178643e2684ba686",
-    "cmd/bucket-metadata-sys.go": "6d0936f537f104bdae1c460554c5ef6f34e6f057d8eb16569a23f012a9498d60",
-    "cmd/bucket-metadata.go": "2b692e42df36373fb98a74792098e1251ae886f5e2759add91a00909b596d04c",
-    "cmd/bucket-policy-handlers.go": "1ac487ef97f63ab19358212f1d582659b4d6fbd17888f5bb21cd5e5028e783e9",
-    "cmd/bucket-versioning-handler.go": "23ec85d45bacf687ada30ea6a06cb5e6d5d5dd32a81865a24aa52a030759e3ce",
-    "cmd/common-main.go": "c06df6c3051ceb264f7ddbdddea33ebe38a59e0c113b5c4c44906c1c2e08a01b",
-    "cmd/erasure-server-pool.go": "1ebb616ac971a22bd736d1b381845c0011cbc95b18dc273e4a946eec8244c80d",
-    "cmd/site-replication-metadata-gate_test.go": "1f170b84d2d25d67a4386451ce26597a752f96c9215155f3156ac1fd140ead86",
-    "cmd/site-replication-metadata-heal_test.go": "37191261f643dae88fbe5be44da8f08b2c52455245d5c61a2df564ecb7c8bffd",
-    "cmd/site-replication-metadata.go": "5756a5495723a2d9456b1c3ed7875cec789298eefb8230ea57126f9de83c4287",
-    "cmd/site-replication-metadata_test.go": "2898c88d48d9693d1c974ee7fd6846b86fe354f378caa630eac2ad0d9b9edfe8",
-    "cmd/site-replication.go": "5015550411bd4d82119cdb96b7f663fd082e4906baee259b30aedbaa616354c8",
-    "cmd/site-replication_test.go": "d85dd2aa991a6cce1fccfa17205e4a478f4ced00e0585fbe2aee2dbf97884b8a",
-    "docs/site-replication/README.md": "abe39529bf5a864dc7c945fd368b647f3fa2121fb324cea1cf63373eff9683a7"
-  },
-  "original_review_head": "4089113e3edbd21a29be8e6af74662462bcc22af",
-  "equivalent_signed_commit": "1ee64a8d895f0876580d5e27d0ce6facf41c347e",
-  "reviewed_tree": "0d18c940c6dec380dffc977cb5720b4e2384e46a",
-  "environment": "go version go1.27.1 darwin/arm64",
-  "scope": "Historical analysis and plan, final review reports and invocation metadata, selected executed test logs, source/binary identity, and rerunnable loopback lab. Excludes model reasoning streams, binaries, temporary volumes and invalid fixture logs. Paths, historical document links and trailing console whitespace were normalized; JSON billing fields and build-info dependency lists were omitted. Original and archived hashes distinguish these editorial transformations.",
-  "final_verdict": "GO_WITH_NONBLOCKING_NOTES",
-  "blocking_findings": 0,
-  "records": [
-    {
-      "file": "issue-77-current.md",
-      "source_record": "issue-77-current.md",
-      "original_sha256": "702f9fe93703c8d4b79a3e2c4709e7f49614dc20211ed5221f3adbd5e3809e2d",
-      "archived_sha256": "b22b50e584de73cfe2f101d35bb912d314feb7f0e0ce48018e1b4a48dc4cf5c5"
-    },
-    {
-      "file": "issue-77-plan.md",
-      "source_record": "issue-77-plan.md",
-      "original_sha256": "2ee214694f16e5949ed58290364770149aee3748faee3614122fee189ef83add",
-      "archived_sha256": "36ae4f99bbc22827cc9a1a437b413713e61f442a568347eeef836b4245c67081"
-    },
-    {
-      "file": "issue-77/current-tests.log",
-      "source_record": "issue-77/current-tests.log",
-      "original_sha256": "d6ed0ef373a37f8254d387a0757a431c2785b6b9a30c4e06d0034688ca3a41e8",
-      "archived_sha256": "d6ed0ef373a37f8254d387a0757a431c2785b6b9a30c4e06d0034688ca3a41e8"
-    },
-    {
-      "file": "issue-77/issue77_heal_review_test.go.txt",
-      "source_record": "issue-77/issue77_heal_review_test.go.txt",
-      "original_sha256": "f1b164afff783282253660b8fa5c131b0de946a078d370cdedd95c338fa73810",
-      "archived_sha256": "f1b164afff783282253660b8fa5c131b0de946a078d370cdedd95c338fa73810"
-    },
-    {
-      "file": "issue-77/issue77_review_test.go.txt",
-      "source_record": "issue-77/issue77_review_test.go.txt",
-      "original_sha256": "77e557bd098fad1345085b6bca27ff8a733984092c68c6b26d0dc74e5fca8caa",
-      "archived_sha256": "77e557bd098fad1345085b6bca27ff8a733984092c68c6b26d0dc74e5fca8caa"
-    },
-    {
-      "file": "issue-77/review/adoption-baseline-probe.go.txt",
-      "source_record": "issue-77/review/adoption-baseline-probe.go.txt",
-      "original_sha256": "4cf7aae6763f1ae736af087244dae01e32af1c546923fa782891759967981574",
-      "archived_sha256": "4cf7aae6763f1ae736af087244dae01e32af1c546923fa782891759967981574"
-    },
-    {
-      "file": "issue-77/review/adoption-baseline-probe.log",
-      "source_record": "issue-77/review/adoption-baseline-probe.log",
-      "original_sha256": "d5ad1bfe7d4faafc7bfc6f55800dc47a7d60b7ec7124cc31ec9ed11b0154bde7",
-      "archived_sha256": "d5ad1bfe7d4faafc7bfc6f55800dc47a7d60b7ec7124cc31ec9ed11b0154bde7"
-    },
-    {
-      "file": "issue-77/review/adoption-baseline-probe.metadata.json",
-      "source_record": "issue-77/review/adoption-baseline-probe.metadata.json",
-      "original_sha256": "618e81cb5852b08fbc3029d812647b9d2dbe3560349ee2e44b860356e12cffa9",
-      "archived_sha256": "a462ad6b10cb7e2737330eed5d5e0a5f98ac0d664aa6fcf40a1f5f2f073709f6"
-    },
-    {
-      "file": "issue-77/review/author-followup-findings.md",
-      "source_record": "issue-77/review/author-followup-findings.md",
-      "original_sha256": "b6d60d813702db300313430325046aa89e45071d3cdc20083d82f061b75ab2c4",
-      "archived_sha256": "b6d60d813702db300313430325046aa89e45071d3cdc20083d82f061b75ab2c4"
-    },
-    {
-      "file": "issue-77/review/decisions-v2.md",
-      "source_record": "issue-77/review/decisions-v2.md",
-      "original_sha256": "52c85f0b45073471e96abe86aa5a182858ff00a5dfe00f290f570f2aad86c5a1",
-      "archived_sha256": "52c85f0b45073471e96abe86aa5a182858ff00a5dfe00f290f570f2aad86c5a1"
-    },
-    {
-      "file": "issue-77/review/decisions-v3.md",
-      "source_record": "issue-77/review/decisions-v3.md",
-      "original_sha256": "7325864c0a3e93689989612da0427b88f890ef8baf197f609c570636b16a2b85",
-      "archived_sha256": "7325864c0a3e93689989612da0427b88f890ef8baf197f609c570636b16a2b85"
-    },
-    {
-      "file": "issue-77/review/decisions-v4.md",
-      "source_record": "issue-77/review/decisions-v4.md",
-      "original_sha256": "982b2e802aed883f2b5e211f0e1677388571ccde8e32af878a35b4192c1f4d16",
-      "archived_sha256": "982b2e802aed883f2b5e211f0e1677388571ccde8e32af878a35b4192c1f4d16"
-    },
-    {
-      "file": "issue-77/review/final-review.md",
-      "source_record": "issue-77/review/final-review.md",
-      "original_sha256": "2410f7b92e2c74b235d54bc155b822675c6b5f9ab0b7226fccca42dc38b0a7ea",
-      "archived_sha256": "2410f7b92e2c74b235d54bc155b822675c6b5f9ab0b7226fccca42dc38b0a7ea"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v1.md",
-      "source_record": "issue-77/review/opus5-max-v1.md",
-      "original_sha256": "8a3b750a261ef520acbe658395d3303cea0835b07184b96022781a2a5d0d4dcf",
-      "archived_sha256": "8a3b750a261ef520acbe658395d3303cea0835b07184b96022781a2a5d0d4dcf"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v1.metadata.json",
-      "source_record": "issue-77/review/opus5-max-v1.metadata.json",
-      "original_sha256": "959a9eac3a8449ef27027fbc65aca55a32b1f46063175d7187922296ecd0a957",
-      "archived_sha256": "d5e3be5c5377adeb88fdcb444ad5b6ee338af91d095027e1c9502298eda51d6d"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v2.md",
-      "source_record": "issue-77/review/opus5-max-v2.md",
-      "original_sha256": "848a476f18ce88208b72d8bf6cb49a2316b2b14dcba64ad66e636fe0a3d70595",
-      "archived_sha256": "848a476f18ce88208b72d8bf6cb49a2316b2b14dcba64ad66e636fe0a3d70595"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v2.metadata.json",
-      "source_record": "issue-77/review/opus5-max-v2.metadata.json",
-      "original_sha256": "ff63843a993830ea48e1a44987390b94613dea99a796453bc91bebca2f4e119f",
-      "archived_sha256": "3cd83d43f461fb1e13bbc4f4163769fa6daf4d151d8442d8615a3e813da38853"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v3.md",
-      "source_record": "issue-77/review/opus5-max-v3.md",
-      "original_sha256": "557ef70288a9a3939842a2f76d63694e2c103a8bfcaa99b7c161ee89e1f58061",
-      "archived_sha256": "557ef70288a9a3939842a2f76d63694e2c103a8bfcaa99b7c161ee89e1f58061"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v3.metadata.json",
-      "source_record": "issue-77/review/opus5-max-v3.metadata.json",
-      "original_sha256": "ad488156f42e33e47ed5d0a267ee69ee82efff69881d0e6400e88cf88a09ef60",
-      "archived_sha256": "1c84da615bccc2f1d7a2aa917ef7af751b0943cc7e413b39f59164a65acc2c98"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v4.md",
-      "source_record": "issue-77/review/opus5-max-v4.md",
-      "original_sha256": "cf10b9d7963ff2b7d4c0fbb1fe744a0e853549eaf7ce675a94cc000f74fe6682",
-      "archived_sha256": "cf10b9d7963ff2b7d4c0fbb1fe744a0e853549eaf7ce675a94cc000f74fe6682"
-    },
-    {
-      "file": "issue-77/review/opus5-max-v4.metadata.json",
-      "source_record": "issue-77/review/opus5-max-v4.metadata.json",
-      "original_sha256": "95bae4427322fc42974a9d94ed5dd6d272a46b6eb669f95669a2fe1ef6ee4d42",
-      "archived_sha256": "60d68b2db11557ee09d60b67748844a56d7b1f323803163ef03d420ad1cc8c42"
-    },
-    {
-      "file": "issue-77/review/plan-v1.md",
-      "source_record": "issue-77/review/plan-v1.md",
-      "original_sha256": "0218a2ca59750acd6a24dc5ece1d55a19762a863f1a2de8edc6ffc9b4ab072e0",
-      "archived_sha256": "97d88bbd010cbd19744c2e2fee0fed8d3713ca553b5d0f06161299772ea8104b"
-    },
-    {
-      "file": "issue-77/review/plan-v2.md",
-      "source_record": "issue-77/review/plan-v2.md",
-      "original_sha256": "2b2c712b17f757ca1a8501714c6c2472bd442a6be50b29d92f2d87e23c412ba8",
-      "archived_sha256": "b193ee6e47b24576f3f4d06608f1bcd0f7f2b0c27ef782ea4f74ee3706f7aa4d"
-    },
-    {
-      "file": "issue-77/review/plan-v3.md",
-      "source_record": "issue-77/review/plan-v3.md",
-      "original_sha256": "8c4ab51a397246a07bb1cc967df3377d26ac084b7928191eb8fecacf6e7c3aed",
-      "archived_sha256": "fb3b37ab9dd2f4ab12b8050293603ed589645a730f1e23c81e2f7162ca46f7d2"
-    },
-    {
-      "file": "issue-77/review/plan-v4.md",
-      "source_record": "issue-77/review/plan-v4.md",
-      "original_sha256": "fbfd37498ab32a589d06254f32182869d5d49fc97722416c46c10b41e5f77816",
-      "archived_sha256": "8624e9b86d3cb6daa497d43ac8cb50318ebabdd7eb4bb7c605e0720522a73bd9"
-    },
-    {
-      "file": "issue-77/review/policy-encoding-probe.go.txt",
-      "source_record": "issue-77/review/policy-encoding-probe.go.txt",
-      "original_sha256": "4e07cba70bdc329a0003ca0b6711886a6de22323293e8de3625cdc838ccaf537",
-      "archived_sha256": "4e07cba70bdc329a0003ca0b6711886a6de22323293e8de3625cdc838ccaf537"
-    },
-    {
-      "file": "issue-77/review/wire-state-probe.go.txt",
-      "source_record": "issue-77/review/wire-state-probe.go.txt",
-      "original_sha256": "cafc6e2d482cc8e178ae3e17651410e3070049cd60009f0edd30fbac29e1309c",
-      "archived_sha256": "cafc6e2d482cc8e178ae3e17651410e3070049cd60009f0edd30fbac29e1309c"
-    },
-    {
-      "file": "issue-77/review/wire-state-probe.log",
-      "source_record": "issue-77/review/wire-state-probe.log",
-      "original_sha256": "5077cd264a50cc05fe9dd628a6e98a3b0aaf94e04b65b712563191e11e4ca9f1",
-      "archived_sha256": "5077cd264a50cc05fe9dd628a6e98a3b0aaf94e04b65b712563191e11e4ca9f1"
-    },
-    {
-      "file": "issue-77/implementation-review/opus5-max-review.md",
-      "source_record": "implementation-review/opus5-max-review.md",
-      "original_sha256": "e40dad6e1969898e35c82b7fd24517a5ee714d603a54fe3598b9e69c20922118",
-      "archived_sha256": "e40dad6e1969898e35c82b7fd24517a5ee714d603a54fe3598b9e69c20922118"
-    },
-    {
-      "file": "issue-77/implementation-review/session.json",
-      "source_record": "implementation-review/session.json",
-      "original_sha256": "c347259a0a501242eb787319d43fe16c117d021ea4e418ce74b09b6c9c33ad08",
-      "archived_sha256": "d26e9c3a7f3d0ba3d34de54e95c64e5939b8d5da1d190bb94d80423b6ca79205"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/opus5-max-review.md",
-      "source_record": "implementation-review/round-2/opus5-max-review.md",
-      "original_sha256": "306a856010f7f9bd1cbec39aba83236978f5a811c98f7fc3f8a08217f72de111",
-      "archived_sha256": "306a856010f7f9bd1cbec39aba83236978f5a811c98f7fc3f8a08217f72de111"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/session.json",
-      "source_record": "implementation-review/round-2/session.json",
-      "original_sha256": "64113634f4e8d0525ed17dd74d092c74419d172aa94b501187daae8ebc27e279",
-      "archived_sha256": "4f4dfe9c50b43b64f15fad830c39c116a296c302655d2c718475249521a68f19"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/opus5-max-review.md",
-      "source_record": "implementation-review/round-3/opus5-max-review.md",
-      "original_sha256": "9106938a7a34cfbc6a11d372dd5ffb71ae7417a77a56d10bbed178f539de8fba",
-      "archived_sha256": "9106938a7a34cfbc6a11d372dd5ffb71ae7417a77a56d10bbed178f539de8fba"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/session.json",
-      "source_record": "implementation-review/round-3/session.json",
-      "original_sha256": "28944efc7a7b2c5451beedf0a05202ffc23b29fad2d9bd04cc89fad3d72184b9",
-      "archived_sha256": "bf31d2dbd8009f0a966c0e820dfdbd5a2820e8b36e71e615f783932fd4ecc9de"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/confirmed-before.log",
-      "source_record": "implementation-review/round-2/confirmed-before.log",
-      "original_sha256": "af2df1c296c3377edb6aae45caa9fafd5e620ffb5a09438c15ca50e3d2cb1c10",
-      "archived_sha256": "af2df1c296c3377edb6aae45caa9fafd5e620ffb5a09438c15ca50e3d2cb1c10"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/confirmed-after.log",
-      "source_record": "implementation-review/round-2/confirmed-after.log",
-      "original_sha256": "0c8be7325817049afe2bd2d3ed668e61cbde2887a443b988b8defbf12304b5bc",
-      "archived_sha256": "0c8be7325817049afe2bd2d3ed668e61cbde2887a443b988b8defbf12304b5bc"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/confirmed-after-race.log",
-      "source_record": "implementation-review/round-2/confirmed-after-race.log",
-      "original_sha256": "da2725133b6c1b41a50a92700ff6adaf6dae7e9ac9d6b5812844715fd2b70e55",
-      "archived_sha256": "da2725133b6c1b41a50a92700ff6adaf6dae7e9ac9d6b5812844715fd2b70e55"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/confirmed-diagnostics-before.log",
-      "source_record": "implementation-review/round-2/confirmed-diagnostics-before.log",
-      "original_sha256": "3d2c1325f1612c9ce79172b148d1ae02cf5443087270e9caeb46046fec1ea2c8",
-      "archived_sha256": "3d2c1325f1612c9ce79172b148d1ae02cf5443087270e9caeb46046fec1ea2c8"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/no-source-before.log",
-      "source_record": "implementation-review/round-2/no-source-before.log",
-      "original_sha256": "e6bac0d117eb195b0ff67ef55eaff302c5f7ee36abf236fdcb69486330b31185",
-      "archived_sha256": "e6bac0d117eb195b0ff67ef55eaff302c5f7ee36abf236fdcb69486330b31185"
-    },
-    {
-      "file": "issue-77/implementation-review/round-2/initial-sync-before.log",
-      "source_record": "implementation-review/round-2/initial-sync-before.log",
-      "original_sha256": "f200f9b961baf20e48f22b50cebd82525dd4647ea42203179d140b5474a28427",
-      "archived_sha256": "f200f9b961baf20e48f22b50cebd82525dd4647ea42203179d140b5474a28427"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/docs-check.json",
-      "source_record": "implementation-review/round-3/docs-check.json",
-      "original_sha256": "76747f4d197d250e6bd3aaa1eba5a7013e56b655f10d143b984faed9d22b503b",
-      "archived_sha256": "322f206ca48816c415ba29f9c67935c0eba038823eddfac8fb7e298ff50bda4a"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/docs-check.log",
-      "source_record": "implementation-review/round-3/docs-check.log",
-      "original_sha256": "af8752f3d8a072954dc6de2b8c72a4054cffbe3583aeae225a8d6623fc95e9fa",
-      "archived_sha256": "a1d67bf392e45da04af9d819a5d76cec98006d1fc78c1a4cce486eb305bdc3ac"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-build.json",
-      "source_record": "implementation-review/round-3/final-build.json",
-      "original_sha256": "924df2a034af81acfe8961489664fa213791096db637e962ced6b78ba7ea8388",
-      "archived_sha256": "c5de13eceee16b23886290fbe351a5269735a549723ed400ddb3e5f0bf7c88e5"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-build.log",
-      "source_record": "implementation-review/round-3/final-build.log",
-      "original_sha256": "d5d78e4d1a5b0dc5bc6042ae94ad64aaf2c500ff3e9091ea1dc9dbdfff04b263",
-      "archived_sha256": "b212590ec72fc2489ea00c36458b3ce900b5b23543c7b141d211ac7b4b06f07b"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-cmd.json",
-      "source_record": "implementation-review/round-3/final-cmd.json",
-      "original_sha256": "431f663b748e81b9c311fa1b931483fe0b1ed09dec796b0894bb5d4a37f5f6b9",
-      "archived_sha256": "39a02d590c9615721fbded8e95cc3b04ebacb32ba5c943d8e9a6f55d912c80c7"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-cmd.log",
-      "source_record": "implementation-review/round-3/final-cmd.log",
-      "original_sha256": "681576adac6d627140a791a22085fc520b61f4a6e04172bc7fb5d0f3967e9c40",
-      "archived_sha256": "67723bedd3472d22a2e3f043e48fdab9bc5e8a212d932c513d44d61cc13eb09d"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-lint.json",
-      "source_record": "implementation-review/round-3/final-lint.json",
-      "original_sha256": "1dc2b38de9c4f37983e45e11ad5f416810f0c96e2cfeeff64b25846544918751",
-      "archived_sha256": "35229aa6e418c54072abf141b4296be491a43cdba87c2f56f853076d78d66559"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-lint.log",
-      "source_record": "implementation-review/round-3/final-lint.log",
-      "original_sha256": "40eb5ec64ae64b73489fc7b6e145aabfea6554e0c73fa22be7b4a8201bb63d18",
-      "archived_sha256": "d4b9455691ff7bcce76a0d886f603eda2179e7fa4e408b8e3bb987d693987c3a"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-runtime.json",
-      "source_record": "implementation-review/round-3/final-runtime.json",
-      "original_sha256": "6f76d83dbf9749f9dde0d6386f0b8a6b52cda6677c17ef9d739f05ea9379c52f",
-      "archived_sha256": "b57c593e1023a0226e7950c142d5bd9a231832a9e7aefc7d9d584075b70b9aca"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-runtime.log",
-      "source_record": "implementation-review/round-3/final-runtime.log",
-      "original_sha256": "4c0137bbe554426026e1d4c8baf48e28368e2450e886c506e0055e0591cd5e5e",
-      "archived_sha256": "2b6228f3c6003e896193be17a7c94970992c1fbcbaba013b1984348bf772bc1f"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-style-target-race.json",
-      "source_record": "implementation-review/round-3/final-style-target-race.json",
-      "original_sha256": "09b77aeb0d720de59ea5b1c31f017e8bf32e89c8dc0d4ecc3f35704f21bc0863",
-      "archived_sha256": "075c8d7393164e5b43364d25c163f5da8675f5b69c67e7523a579943a710aaca"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-style-target-race.log",
-      "source_record": "implementation-review/round-3/final-style-target-race.log",
-      "original_sha256": "3089ad9a1c7283a23853a137311728598a7e1ae3bdbbf16e5f000e1149a7e12f",
-      "archived_sha256": "a2609e9a889bcc68828e58c15967bb04820a92462c3f94de86ea584e33a6660a"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-target-race.json",
-      "source_record": "implementation-review/round-3/final-target-race.json",
-      "original_sha256": "68dfcad5fc63d5981065f86fb00482cc466c46f275c4c340095c26ad719f079c",
-      "archived_sha256": "5fb25fa150e66478b2d170faeed3374d71325311f7e625996fe1653cfc957b66"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-target-race.log",
-      "source_record": "implementation-review/round-3/final-target-race.log",
-      "original_sha256": "b437b206ebd12fff304f9cdb59b39f25fd0ecf54e9f510aca98c444b0b516c89",
-      "archived_sha256": "a5b6b6337ebbcf0eca02c2ca15bb1fa562d39ab6fe29f41cd304dd2979976f59"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-tree-equivalence.json",
-      "source_record": "implementation-review/round-3/final-tree-equivalence.json",
-      "original_sha256": "f7d1563d2f675ae282b4bec0bcedcc9809b0a0c985332d862c61ff4b8c75a59f",
-      "archived_sha256": "f7d1563d2f675ae282b4bec0bcedcc9809b0a0c985332d862c61ff4b8c75a59f"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-vet.json",
-      "source_record": "implementation-review/round-3/final-vet.json",
-      "original_sha256": "934ae2812aef8b8635517f838a07cc81dc4780603d621770f9a150bf00d43561",
-      "archived_sha256": "be1b653811da4f6503f329f7e7c8b321b3426107548f0079c8560b177bc0c37f"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/final-vet.log",
-      "source_record": "implementation-review/round-3/final-vet.log",
-      "original_sha256": "f8166088e218c7c8afa25988c02cee77ec58b552f531121a07959b31bd97ebbb",
-      "archived_sha256": "012d45737551ed5ffdccc3b3535ce0be152a9e9f869f50779d5da04e965a9936"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/test-style.diff",
-      "source_record": "implementation-review/round-3/test-style.diff",
-      "original_sha256": "69a7f7efbb17d8dcdd2a0d7f91098569a48d7f1aeeecb6d2aefe04b6047a9025",
-      "archived_sha256": "69a7f7efbb17d8dcdd2a0d7f91098569a48d7f1aeeecb6d2aefe04b6047a9025"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-crosscompile-after.log",
-      "source_record": "implementation-review/ci-crosscompile-after.log",
-      "original_sha256": "68ef16136fabb16babd8fa26d2e3fcf6ffddac30002f874f7b9dfdd61961e82f",
-      "archived_sha256": "68ef16136fabb16babd8fa26d2e3fcf6ffddac30002f874f7b9dfdd61961e82f"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-gen-after.log",
-      "source_record": "implementation-review/ci-gen-after.log",
-      "original_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-      "archived_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-internal-after.log",
-      "source_record": "implementation-review/ci-internal-after.log",
-      "original_sha256": "775e5b2b35fda828ffd69f47902e065f58aa2e2ecd915ace9e50f4085aa2d0c4",
-      "archived_sha256": "775e5b2b35fda828ffd69f47902e065f58aa2e2ecd915ace9e50f4085aa2d0c4"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-lint-after.log",
-      "source_record": "implementation-review/ci-lint-after.log",
-      "original_sha256": "cd29668fa61e94079dda874e10920623279083c491798fa9cbdb903ebeee4d82",
-      "archived_sha256": "cd29668fa61e94079dda874e10920623279083c491798fa9cbdb903ebeee4d82"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-s3select-after.log",
-      "source_record": "implementation-review/ci-s3select-after.log",
-      "original_sha256": "30fc4888ba394ef0daa6c76fa497c28086b7307ec30d195fe4f2b725badc59c9",
-      "archived_sha256": "30fc4888ba394ef0daa6c76fa497c28086b7307ec30d195fe4f2b725badc59c9"
-    },
-    {
-      "file": "issue-77/implementation-review/ci-verify-after.log",
-      "source_record": "implementation-review/ci-verify-after.log",
-      "original_sha256": "2b6860f20bdfd453bcc299f8b57a117cd698bab77d6a474913e31794c3937bd3",
-      "archived_sha256": "2b6860f20bdfd453bcc299f8b57a117cd698bab77d6a474913e31794c3937bd3"
-    },
-    {
-      "file": "issue-77/implementation-review/round-3/binary-identity.json",
-      "source_record": "implementation-review/round-3/binary-identity.json",
-      "original_sha256": "ffdb768a1931388f174eaca0980cc840015a8e0995b5509ef68718d10044f5e5",
-      "archived_sha256": "d4ea7abc1f419e834746820b5b2d2f1b3264e42114161e7c6803f4e952fcf17d"
-    },
-    {
-      "file": "issue-77/runtime/converged.json",
-      "source_record": "runtime-reviewed/converged.json",
-      "original_sha256": "d6d97bc01a12fd624af37432e9558356255b5cd9d4333eaa616e39998c2ce94f",
-      "archived_sha256": "d6d97bc01a12fd624af37432e9558356255b5cd9d4333eaa616e39998c2ce94f"
-    },
-    {
-      "file": "issue-77/runtime/main.go",
-      "source_record": "twosite/main.go",
-      "original_sha256": "ec63a5e2553907a826727af416145057215adabfb9fa2839861f0c05b7cc2a8f",
-      "archived_sha256": "023459a3873753d5a373c2ba6c078d690c9a2a18c31efa98ecd5f4976ecaba4a"
-    }
-  ]
-}
@@ -1,129 +0,0 @@
-=== RUN   TestQueuedMetadataUpdateAfterDelete
-=== RUN   TestQueuedMetadataUpdateAfterDelete/expiry=false
-=== RUN   TestQueuedMetadataUpdateAfterDelete/expiry=true
--- PASS: TestQueuedMetadataUpdateAfterDelete (0.43s)
-    --- PASS: TestQueuedMetadataUpdateAfterDelete/expiry=false (0.22s)
-    --- PASS: TestQueuedMetadataUpdateAfterDelete/expiry=true (0.21s)
-=== RUN   TestIssue77CurrentHealInputs
-=== RUN   TestIssue77CurrentHealInputs/ErasureSD/quota-tombstone-cache
-    issue77_heal_review_test.go:58: QUOTA_CACHE: disk has no quota, cache still enforces 1024
-=== RUN   TestIssue77CurrentHealInputs/ErasureSD/same-payload-time-barrier
-    issue77_heal_review_test.go:79: BARRIER_NOT_HEALED: same payload remains at 2026-09-12 06:59:55.50105 +0800 CST, latest is 2026-09-11 23:00:55.50105 +0000 UTC
-=== RUN   TestIssue77CurrentHealInputs/ErasureSD/creation-default-selection
-    issue77_heal_review_test.go:105: DEFAULT_SELECTED: creation-default erased valid policy in 5/32 heal rounds
-=== RUN   TestIssue77CurrentHealInputs/ErasureSD/tag-remote-source-time
-    issue77_heal_review_test.go:144: TAG_WIRE_TIME: sent 0001-01-01 00:00:00 +0000 UTC, want 2026-09-11 23:00:55.50105 +0000 UTC
-=== RUN   TestIssue77CurrentHealInputs/Erasure/quota-tombstone-cache
-    issue77_heal_review_test.go:58: QUOTA_CACHE: disk has no quota, cache still enforces 1024
-=== RUN   TestIssue77CurrentHealInputs/Erasure/same-payload-time-barrier
-    issue77_heal_review_test.go:79: BARRIER_NOT_HEALED: same payload remains at 2026-09-12 06:59:55.71847 +0800 CST, latest is 2026-09-11 23:00:55.71847 +0000 UTC
-=== RUN   TestIssue77CurrentHealInputs/Erasure/creation-default-selection
-    issue77_heal_review_test.go:105: DEFAULT_SELECTED: creation-default erased valid policy in 2/32 heal rounds
-=== RUN   TestIssue77CurrentHealInputs/Erasure/tag-remote-source-time
-    issue77_heal_review_test.go:144: TAG_WIRE_TIME: sent 0001-01-01 00:00:00 +0000 UTC, want 2026-09-11 23:00:55.71847 +0000 UTC
--- FAIL: TestIssue77CurrentHealInputs (1.21s)
-    --- FAIL: TestIssue77CurrentHealInputs/ErasureSD/quota-tombstone-cache (0.01s)
-    --- FAIL: TestIssue77CurrentHealInputs/ErasureSD/same-payload-time-barrier (0.00s)
-    --- FAIL: TestIssue77CurrentHealInputs/ErasureSD/creation-default-selection (0.08s)
-    --- FAIL: TestIssue77CurrentHealInputs/ErasureSD/tag-remote-source-time (0.00s)
-    --- FAIL: TestIssue77CurrentHealInputs/Erasure/quota-tombstone-cache (0.04s)
-    --- FAIL: TestIssue77CurrentHealInputs/Erasure/same-payload-time-barrier (0.02s)
-    --- FAIL: TestIssue77CurrentHealInputs/Erasure/creation-default-selection (0.83s)
-    --- FAIL: TestIssue77CurrentHealInputs/Erasure/tag-remote-source-time (0.02s)
-=== RUN   TestIssue77CurrentPeerCheckBeforeLock
-=== RUN   TestIssue77CurrentPeerCheckBeforeLock/ErasureSD
-    issue77_heal_review_test.go:209: CHECK_OUTSIDE_LOCK: queued older event replaced newer committed tags with "<Tagging><TagSet><Tag><Key>key</Key><Value>older</Value></Tag></TagSet></Tagging>"
-=== RUN   TestIssue77CurrentPeerCheckBeforeLock/Erasure
-    issue77_heal_review_test.go:209: CHECK_OUTSIDE_LOCK: queued older event replaced newer committed tags with "<Tagging><TagSet><Tag><Key>key</Key><Value>older</Value></Tag></TagSet></Tagging>"
--- FAIL: TestIssue77CurrentPeerCheckBeforeLock (0.30s)
-    --- FAIL: TestIssue77CurrentPeerCheckBeforeLock/ErasureSD (0.01s)
-    --- FAIL: TestIssue77CurrentPeerCheckBeforeLock/Erasure (0.07s)
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/policy
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.017159 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/tags
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.029759 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.03401 +0800 CST
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/sse
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.04395 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.048028 +0800 CST
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/quota
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.057746 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.061453 +0800 CST
-    issue77_review_test.go:112: older source PUT did not resurrect config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/versioning
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.069105 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:87: nil payload is correctly a no-op
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/objectlock
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.077601 +0800 CST, want source 2026-09-11 22:59:57.012479 +0000 UTC
-    issue77_review_test.go:87: nil payload is correctly a no-op
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/policy
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.247526 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/tags
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.369557 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.394147 +0800 CST
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/sse
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.466111 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.489771 +0800 CST
-    issue77_review_test.go:110: STALE_RESURRECTION: older source PUT resurrected a locally deleted config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/quota
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.562629 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:91: NEWER_DELETE: HTTP 200, but live config remained after newer source DELETE
-    issue77_review_test.go:105: TOMBSTONE_EXPORT: got 0001-01-01 00:00:00 +0000 UTC, want 2026-09-12 07:49:57.590096 +0800 CST
-    issue77_review_test.go:112: older source PUT did not resurrect config
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/versioning
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.640941 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:87: nil payload is correctly a no-op
-=== RUN   TestIssue77CurrentSourceTimeAndDeletion/Erasure/objectlock
-    issue77_review_test.go:79: SOURCE_TIME: persisted 2026-09-12 07:49:57.690024 +0800 CST, want source 2026-09-11 22:59:57.221624 +0000 UTC
-    issue77_review_test.go:87: nil payload is correctly a no-op
--- FAIL: TestIssue77CurrentSourceTimeAndDeletion (0.78s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/policy (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/tags (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/sse (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/quota (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/versioning (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/ErasureSD/objectlock (0.01s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/policy (0.12s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/tags (0.10s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/sse (0.10s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/quota (0.08s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/versioning (0.05s)
-    --- FAIL: TestIssue77CurrentSourceTimeAndDeletion/Erasure/objectlock (0.05s)
-=== RUN   TestIssue77CurrentBulkApplyOrdering
-=== RUN   TestIssue77CurrentBulkApplyOrdering/ErasureSD
-    issue77_review_test.go:143: BULK_STALE_OVERWRITE: older bulk event overwrote newer config, value="<Tagging><TagSet><Tag><Key>key</Key><Value>old</Value></Tag></TagSet></Tagging>" time=2026-09-12 08:48:57.792608 +0800 CST
-=== RUN   TestIssue77CurrentBulkApplyOrdering/Erasure
-    issue77_review_test.go:143: BULK_STALE_OVERWRITE: older bulk event overwrote newer config, value="<Tagging><TagSet><Tag><Key>key</Key><Value>old</Value></Tag></TagSet></Tagging>" time=2026-09-12 08:48:57.880379 +0800 CST
--- FAIL: TestIssue77CurrentBulkApplyOrdering (0.27s)
-    --- FAIL: TestIssue77CurrentBulkApplyOrdering/ErasureSD (0.01s)
-    --- FAIL: TestIssue77CurrentBulkApplyOrdering/Erasure (0.05s)
-=== RUN   TestPeerBucketAdoptionPreservesLockAndVersioningConfigs
--- PASS: TestPeerBucketAdoptionPreservesLockAndVersioningConfigs (0.30s)
-=== RUN   TestPeerBucketAdoptionBootstrapsMissingConfigs
--- PASS: TestPeerBucketAdoptionBootstrapsMissingConfigs (0.24s)
-=== RUN   TestPeerBucketAdoptionNormalizesVersioningWhenEnablingLock
--- PASS: TestPeerBucketAdoptionNormalizesVersioningWhenEnablingLock (0.28s)
-=== RUN   TestPeerBucketAdoptionEnablesSuspendedVersioning
--- PASS: TestPeerBucketAdoptionEnablesSuspendedVersioning (0.27s)
-=== RUN   TestPeerBucketObjectLockMetadataCurrentAndLegacyPayloads
--- PASS: TestPeerBucketObjectLockMetadataCurrentAndLegacyPayloads (0.30s)
-=== RUN   TestPeerBucketObjectLockMetadataWithoutLockEnabled
--- PASS: TestPeerBucketObjectLockMetadataWithoutLockEnabled (0.24s)
-=== RUN   TestSiteReplicationStatusAccountsPerSiteAndSurvivesMalformedConfig
--- PASS: TestSiteReplicationStatusAccountsPerSiteAndSurvivesMalformedConfig (0.52s)
-FAIL
-FAIL	github.com/minio/minio/cmd	7.167s
-FAIL
@@ -1,7 +0,0 @@
-Testing builds for OS/Arch: linux/amd64 linux/arm64 darwin/amd64 darwin/arm64 windows/amd64 windows/arm64
-->     linux/amd64:github.com/minio/minio
-->     linux/arm64:github.com/minio/minio
-->    darwin/amd64:github.com/minio/minio
-->    darwin/arm64:github.com/minio/minio
-->   windows/amd64:github.com/minio/minio
-->   windows/arm64:github.com/minio/minio
@@ -1,78 +0,0 @@
-ok  	github.com/minio/minio/internal/amztime	5.149s
-ok  	github.com/minio/minio/internal/arn	13.830s
-ok  	github.com/minio/minio/internal/auth	5.472s
-ok  	github.com/minio/minio/internal/bpool	13.244s
-ok  	github.com/minio/minio/internal/bucket/bandwidth	12.671s
-ok  	github.com/minio/minio/internal/bucket/cors	2.705s
-ok  	github.com/minio/minio/internal/bucket/encryption	12.228s
-ok  	github.com/minio/minio/internal/bucket/lifecycle	3.223s
-ok  	github.com/minio/minio/internal/bucket/object/lock	11.599s
-ok  	github.com/minio/minio/internal/bucket/replication	12.817s
-ok  	github.com/minio/minio/internal/bucket/versioning	13.157s
-ok  	github.com/minio/minio/internal/cachevalue	18.526s
-?   	github.com/minio/minio/internal/color	[no test files]
-ok  	github.com/minio/minio/internal/config	14.365s
-?   	github.com/minio/minio/internal/config/api	[no test files]
-?   	github.com/minio/minio/internal/config/batch	[no test files]
-?   	github.com/minio/minio/internal/config/browser	[no test files]
-?   	github.com/minio/minio/internal/config/callhome	[no test files]
-ok  	github.com/minio/minio/internal/config/compress	15.754s
-ok  	github.com/minio/minio/internal/config/dns	17.351s
-?   	github.com/minio/minio/internal/config/drive	[no test files]
-ok  	github.com/minio/minio/internal/config/etcd	18.460s
-?   	github.com/minio/minio/internal/config/heal	[no test files]
-ok  	github.com/minio/minio/internal/config/identity/ldap	19.132s
-ok  	github.com/minio/minio/internal/config/identity/openid	19.829s
-?   	github.com/minio/minio/internal/config/identity/openid/provider	[no test files]
-?   	github.com/minio/minio/internal/config/identity/plugin	[no test files]
-?   	github.com/minio/minio/internal/config/identity/tls	[no test files]
-ok  	github.com/minio/minio/internal/config/ilm	19.256s
-?   	github.com/minio/minio/internal/config/lambda	[no test files]
-ok  	github.com/minio/minio/internal/config/lambda/event	18.225s
-?   	github.com/minio/minio/internal/config/lambda/target	[no test files]
-ok  	github.com/minio/minio/internal/config/notify	17.465s
-?   	github.com/minio/minio/internal/config/policy/opa	[no test files]
-?   	github.com/minio/minio/internal/config/policy/plugin	[no test files]
-?   	github.com/minio/minio/internal/config/scanner	[no test files]
-ok  	github.com/minio/minio/internal/config/storageclass	17.289s
-ok  	github.com/minio/minio/internal/config/subnet	17.168s
-ok  	github.com/minio/minio/internal/crypto	18.081s
-ok  	github.com/minio/minio/internal/deadlineconn	21.703s
-ok  	github.com/minio/minio/internal/disk	17.514s
-ok  	github.com/minio/minio/internal/dsync	154.956s
-ok  	github.com/minio/minio/internal/etag	17.755s
-ok  	github.com/minio/minio/internal/event	17.598s
-ok  	github.com/minio/minio/internal/event/target	16.108s
-ok  	github.com/minio/minio/internal/grid	22.027s
-ok  	github.com/minio/minio/internal/handlers	14.030s
-ok  	github.com/minio/minio/internal/hash	14.016s
-?   	github.com/minio/minio/internal/hash/sha256	[no test files]
-ok  	github.com/minio/minio/internal/http	13.661s
-?   	github.com/minio/minio/internal/init	[no test files]
-ok  	github.com/minio/minio/internal/ioutil	14.637s
-ok  	github.com/minio/minio/internal/jwt	13.120s
-ok  	github.com/minio/minio/internal/kms	12.536s
-ok  	github.com/minio/minio/internal/lock	13.029s
-ok  	github.com/minio/minio/internal/logger	12.231s
-?   	github.com/minio/minio/internal/logger/message/audit	[no test files]
-?   	github.com/minio/minio/internal/logger/target/console	[no test files]
-?   	github.com/minio/minio/internal/logger/target/http	[no test files]
-?   	github.com/minio/minio/internal/logger/target/kafka	[no test files]
-?   	github.com/minio/minio/internal/logger/target/loggertypes	[no test files]
-?   	github.com/minio/minio/internal/logger/target/testlogger	[no test files]
-ok  	github.com/minio/minio/internal/lsync	22.092s
-?   	github.com/minio/minio/internal/mcontext	[no test files]
-?   	github.com/minio/minio/internal/mountinfo	[no test files]
-?   	github.com/minio/minio/internal/net	[no test files]
-?   	github.com/minio/minio/internal/once	[no test files]
-ok  	github.com/minio/minio/internal/pubsub	11.507s
-ok  	github.com/minio/minio/internal/rest	10.318s
-ok  	github.com/minio/minio/internal/ringbuffer	10.927s
-ok  	github.com/minio/minio/internal/s3select	8.665s
-ok  	github.com/minio/minio/internal/s3select/csv	8.862s
-ok  	github.com/minio/minio/internal/s3select/json	8.547s
-ok  	github.com/minio/minio/internal/s3select/jstream	8.936s
-?   	github.com/minio/minio/internal/s3select/parquet	[no test files]
-?   	github.com/minio/minio/internal/s3select/simdj	[no test files]
-ok  	github.com/minio/minio/internal/s3select/sql	8.637s
-ok  	github.com/minio/minio/internal/store	9.594s
@@ -1,13 +0,0 @@
-Installing golangci-lint v2.13.1
-golangci/golangci-lint info checking GitHub for tag 'v2.13.1'
-golangci/golangci-lint info found version: 2.13.1 for v2.13.1/darwin/arm64
-golangci/golangci-lint info installed .bin/golangci/v2.13.1/golangci-lint
-Running lint check
-0 issues.
-typos binary is not found.. skipping..
-LINT_EXIT=0
-compatibility manifest: imports=119 env=438 metrics=19 headers=87 routes=223 roots=1 grid=3 storage=15 policy=58 brand=181 sha256=db0d00c26360412a5388be1eb87c2dd763633ae243c558015f7b3ae86c80baaf
-Silo rebrand compatibility baseline is unchanged
-Silo delivery and runtime rebrand checks passed
-docker entrypoint argv compatibility tests passed
-REBRAND_EXIT=0
@@ -1,7 +0,0 @@
-ok  	github.com/minio/minio/internal/s3select	2.150s
-ok  	github.com/minio/minio/internal/s3select/csv	3.612s
-ok  	github.com/minio/minio/internal/s3select/json	2.105s
-ok  	github.com/minio/minio/internal/s3select/jstream	2.937s
-?   	github.com/minio/minio/internal/s3select/parquet	[no test files]
-?   	github.com/minio/minio/internal/s3select/simdj	[no test files]
-ok  	github.com/minio/minio/internal/s3select/sql	2.620s
@@ -1,2 +0,0 @@
-Silo delivery and runtime rebrand checks passed
-docker entrypoint argv compatibility tests passed
@@ -1,292 +0,0 @@
-# SILO #77 实现深度对抗性代码审查报告
-
-审查对象 `4089113e3`（基线 `5c5765816`）。只读 worktree，全部结论来自直接阅读 `cmd/` 生产代码、`silo-pkg`/`madmin-go` 依赖源码与调用链，并与证据目录交叉核对。测试 PASS 仅当作"已覆盖的观察"，不作正确性证明。
-
-> 说明：本会话的 Write/ExitPlanMode 工具不可用，报告直接输出于此，未落盘。
-
---
-
-## 0. Verdict
-
-**GO_WITH_NONBLOCKING_NOTES**
-
- **无条件阻断项：0**
- **条件性阻断项：1**（F1）——条件：支持矩阵中存在 `.metadata.bin` 缺失或 `Created == 0` 的桶
- 其余：F2 (S2)、F3 (S3)、F4 (S3)、F5（最小性）、F6（nit）、F7（范围说明）、F8（证据不足）、F9/F10（残留边界与覆盖缺口）
-
-我逐条读了六类配置在全部入口的生产路径，**没有**找到在"同一桶世代、Created 已知、合法带源时间事件"范围内会发生回退、删除复活、锁外判旧或重复广播的反例。核心机制是正确的。缺陷集中在两处边缘：零 Created 历史桶被新代码判为不可写（F1，且其自带测试用 stub 掩盖了生产行为），以及 Policy 规范编码排序与既有公开 mismatch 统计口径冲突（F2）。
-
---
-
-## 1. Findings
-
-### F1 — 零 Created / 无 metadata.bin 的桶：六类配置写入全部硬失败；设计中的"物理桶 Created 补齐"在生产中不可达；配套测试用 stub 掩盖
-
-**严重度 S1（条件性阻断）/ 否则 S2 · 已证实（源码级链路完整），需运行定量确认**
-
-位置：
- `cmd/bucket-metadata-replication.go:265-278` `ensureBucketMetadataCreated`
- `cmd/bucket-metadata-sys.go:181-185`（六类分支入口，失败即 `return err`）
- `cmd/erasure-server-pool.go:2280-2285` —— **`GetBucketInfo` 无条件用缓存 `meta.Created` 覆盖物理卷 Created**
- `cmd/admin-bucket-handlers.go:1092`（import）、`cmd/site-replication.go:1671`（bulk peer apply）
- `cmd/site-replication.go:2156-2166` + `cmd/site-replication-metadata.go:45-55`（初次同步静默跳过）
- `cmd/site-replication-metadata-gate_test.go:145-159`（stub）
-
-最短触发链：
-1. 桶存在，但 `.minio.sys/buckets/<b>/.metadata.bin` 不存在且无 legacy 配置文件 → `loadBucketMetadataParse` 走 `errConfigNotFound` 分支并**返回 nil error 且 `Created == 0`**（`cmd/bucket-metadata.go:230-297`；注意 `defaultTimestamps()` 只在 `err == nil` 时调用）。
-2. 启动时 `concurrentLoad`（`cmd/bucket-metadata-sys.go:737-763`）把这个 `Created == 0` 的 meta 写进 `metadataMap`。
-3. `PUT ?tagging`（或 policy/sse/quota/versioning/object-lock）→ `updateAndParseMetadata` → `ensureBucketMetadataCreated` → `objAPI.GetBucketInfo(..., NoMetadata:true)` → `erasureServerPools.GetBucketInfo` 先取到物理卷时间，**随后被缓存里的 0 覆盖** → `info.Created.IsZero()` → `errors.New("bucket metadata creation time is unknown")`。
-4. 客户端收到错误。基线版本此处成功（直接赋值 + `saveMetadata`）。
-
-注：`getAllLegacyConfigs` 在 `cmd/bucket-metadata.go:468` 会用 `info.ModTime` 填 `Created`，所以**有** legacy 配置文件的桶反而没事；**从未设置过任何桶配置**的老桶才是高发场景。
-
-用户影响：
- 六类配置 PUT/DELETE 全部返回错误，**无运维恢复路径**；`mc admin bucket import` 对该桶整体失败（`rpt.SetStatus(bucket, "", err)`）；peer bulk apply 返回错误导致该桶复制持续报错。
- `syncToAllPeers` 对这类桶**静默跳过全部五类配置**（`state.candidate()` 要求 `valid`，而 `valid = !created.IsZero() && ...`），且**无任何诊断日志** —— 基线版本会发送。
-
-最小修复（推荐 a）：
- **(a)** `cmd/erasure-server-pool.go:2282` 改为 `if !meta.Created.IsZero() { bucketInfo.Created = meta.Created }`。一行；恢复计划中"物理桶 Created 补齐路径"的本意；顺带修掉"无 metadata.bin 的桶在 `mc ls` 里创建时间为零"这一既有瑕疵。需复核 `bucketExists`、`hasBucket = !bi.CreatedAt.IsZero()` 等消费点 —— Created 由零变非零对它们都是变好。
- **(b)** 另外在 `syncToAllPeers`（`cmd/site-replication.go:2138-2156`）把已在手的 `bucketInfo.Created` 兜底进 `meta.Created`，或在"有内容但 send==false"时打一条 indeterminate，避免静默漏发。
-
-应补验证：
- 把 `TestPeerBucketMetadataUnknownCreated` 的 `"physical-created"` 子用例**去掉 stub**：用真实 ObjectLayer，把 `newBucketMetadata(bucket)`（Created 为零）存盘并让缓存持有它，断言 `globalBucketMetadataSys.Update(ctx, bucket, bucketTaggingConfig, tagXML)` 成功且 `Created` 被补齐为物理时间。**当前实现会失败。**
- 一个 `syncToAllPeers` 用例：`Created == 0` 且 Tags 非空的桶，断言初次同步仍发 Tags 事件（或至少有诊断）。
-
---
-
-### F2 — Policy 规范编码对 `Statement` 数组排序，与既有公开统计的顺序敏感 `Equals` 冲突：永久假 mismatch，heal 永不修复
-
-**严重度 S2（公开状态/可观测性，非数据损失）· 已证实（源码级），需运行确认**
-
-位置：
- `cmd/bucket-metadata-replication.go:110` —— `sort.Slice` 作用于**每个**数组，含顶层 `Statement`
- `cmd/bucket-metadata-replication.go:124-161` `canonicalBucketPolicy`
- `cmd/site-replication.go:3796` `isBktPolicyReplicated` → `prev.Equals(*p)`（`silo-pkg .../policy/bucket-policy.go:190-194` 按 `Statements[i]` **下标**逐一比较，顺序敏感）
- `cmd/site-replication-metadata.go:174` vs `:179` —— 本地 heal 写**源站原始字节**，远端 heal 经 `PeerBucketPolicyHandler` → `canonicalBucketPolicy` 写**规范字节**
-
-最短反例：
-1. 站点 A 存在升级前写入的桶策略，statement 顺序不是规范字节序（极常见，例如 Deny 在 Allow 之前）。
-2. 加入新站点 B：`syncToAllPeers` → `initialBucketConfigReplicationEvent` 发送 **A 的原始字节**，`UpdatedAt = A.PolicyConfigUpdatedAt`。
-3. B 的 admin 入口解析后调 `PeerBucketPolicyHandler` → `canonicalBucketPolicy` → 存**排序后**字节。
-4. 此后 A/B 的比较键（canonical）完全相同、时间相同 → `compareBucketConfigStates == 0` → heal 永不写、永不发 RPC；而 `Equals` 因 statement 顺序不同返回 false → `mc admin replicate status` **永久**报 bucket policy mismatch，`ReplicatedBucketPolicies` 少计。
-
-同机制第二条路径：同一轮 heal 中本地目标拿原始字节、远端目标拿规范字节，三站点集群会出现持久分叉。
-
-为何是新问题：基线 `PeerBucketPolicyHandler` 用 `json.Marshal(policy)`，集合（Action/Resource/Principal）顺序随机但 **statement 切片顺序被 marshal/unmarshal 保持**，`Equals` 一直成立。
-
-最小修复（三选一，推荐 a 或 b）：
- (a) `canonicalBucketPolicyJSON` 不对顶层 `Statement` 数组排序（只排集合数组）——顺序在全链路被保留，两站点独立解析同一文档仍得相同键。
- (b) `isBktPolicyReplicated` 改用 `canonicalBucketPolicy` 的键比较，而不是 `Equals`。
- (c) `healBucketConfig` 本地分支改写 `incoming.data`（与远端同一编码）——只修本地/远端分叉，**不**修 A 与新站点的分叉。
-
-应补验证：双站点用例——A 侧直接把 legacy 顺序的策略字节写盘（绕过 canonical 编码器），join B，跑两轮 heal，断言 `SiteReplicationStatus` 不报 policy mismatch。
-
---
-
-### F3 — heal 诊断以 ERROR 级、按"桶 × 字段"输出；正常瞬态也报警，且单一 key 会吞掉真实 RPC 失败
-
-**严重度 S3 · 已证实**
-
-位置：`cmd/site-replication-metadata.go:134-145`（两个诊断循环）、`:182-186`（peer 错误）、`cmd/logging.go:19-21`（`replLogOnceIf` 无 errKind ⇒ **ErrorKind**）
-
-1. 第二个循环对 `!state.valid` 的目标打 `indeterminate / unusable peer`。`valid` 要求对端 `CreatedAt != 0` —— 而**对端还没有这个桶**时 `CreatedAt` 就是 0（`cmd/site-replication.go:3096-3103` 给缺桶站点填零值 `SRBucketInfo`）。于是"桶刚建、尚未传播"这种完全正常的瞬态，对每个桶产生 6 条 **ERROR**。
-2. 第一个循环在对端 `RemoteTargetConnectionErr`（`cmd/site-replication.go:3020-3026` 填空 ID）时，对**每个本地桶 × 6 字段**打 `missing peer`。一个站点掉线 ≈ 每小时 6×桶数 条 ERROR。
-3. 四种完全不同的情况（缺 peer / peer 不可用 / peer RPC 失败 / Created 未知）共用同一个 key `"bucket-metadata/<bucket>/<file>/indeterminate"` **且错误正文相同**，而 `logOnceIf` 只按 key+正文去重（`internal/logger/logonce.go:100-119`）。结果：**真正的 heal RPC 失败可能被同桶同字段的"缺 peer"消息顶掉而完全不打印**；基线版本这些失败走 `replLogIf`，必然打印。
-
-最小修复：`target.CreatedAt.IsZero()`（桶不在对端）与 `found == false` 时不打诊断；把 4 种情况拆成不同 reason（`unreachable`/`peer-error`/`unknown-created`）避免 key 撞车；给 `logBucketConfigReplication` 传 `logger.WarningKind`。
-
---
-
-### F4 — 三处用户可见语义变更未写入文档
-
-**严重度 S3（文档/兼容）· 已证实（逐条比对 `docs/site-replication/README.md:65-118`）**
-
-1. **空 Bucket Policy PUT 现在立即按删除处理**：`cmd/bucket-policy-handlers.go:102` + `cmd/bucket-metadata-replication.go:176-180` → `canonicalBucketPolicy` 对 `IsEmpty()` 返回 nil → 落盘 nil → `GetBucketPolicy` 从"200 + 空策略文档"变为 **404 NotFound**。计划 §1 明确写了"需要写入兼容说明"，文档里没有。
-2. **零 Quota (`{}`) 在对端从"被删除"变为"保留 live 文档"**：`cmd/admin-bucket-handlers.go:79-96` 删除了出站 `bucketMeta.Quota = nil` 改写。这是正确的对齐（本地原本就保留 `{}`），但对端 `GetBucketQuotaConfig` 的结果会变。
-3. **`GET ?policy` 与 `mc admin bucket export` 的 JSON 形态改变**：`cmd/bucket-policy-handlers.go:202`、`cmd/admin-bucket-handlers.go:439` 改用 `canonicalBucketPolicy` → 对象键按字母序、集合数组与 **Statement 数组被排序**。语义等价（S3 策略求值与 statement 顺序无关），但字节级比对输出的工具会看到变化。
-
---
-
-### F5 — 最小性：Policy 规范编码器接入 PUT / GET / export / peer 落盘，并非 #77 不变量所必需
-
-**非缺陷，最小性判断 · 已证实**
-
-比较键在 `bucketConfigPayload`（`cmd/bucket-metadata-replication.go:172-180`）里**从已解析策略现算**，与落盘字节无关。即使 PUT/peer 仍用 `json.Marshal`（字节不确定），两站点的键依旧相同，收敛性完全不受影响。
-
-因此这部分改动带来的是两件**额外**的事：(a) 让原本被 `ActionSet.MarshalJSON`（`silo-pkg .../policy/actionset.go:144-148`，空集合报错）挡掉的 `NotAction`/`NotResource` 策略首次可以写入 —— 这是**功能新增**；(b) 直接导致 F2。
-
-判断：如果作者**有意**支持负集合策略，应作为独立特性声明并单独记录/测试（目前只有 `site-replication-metadata_test.go:530-541` 一个用例）；如果只为 #77，最小做法是 `canonicalBucketPolicy` 只用于比较键。我**不**主张必须删除——它确实修掉了"能存不能读"的潜在坑——但必须承认这是范围外的行为扩张，且未在文档中声明。
-
-其它可删复杂度（都很小，不影响不变量）：
- `cmd/bucket-metadata-sys.go:151` 在零值结构体上用 `replicatedBucketConfig(&result.meta, configFile)` 做"是不是这六类"的判断，语义晦涩；一个 `isReplicatedBucketConfig(file) bool` 更清楚。
- 修掉 F3 后，`healBucketConfig` 的第二个诊断循环（`:139-145`）可并入主循环。
-
---
-
-### F6 — Nit
-
- `isBucketMetadataEqual`（`cmd/site-replication.go:5164`）现在只被测试引用，生产死代码。
- `cmd/bucket-metadata-replication.go:36-38` 注释"Object Lock is applied before Versioning"对 bulk apply 循环成立，但 `healBuckets`（`cmd/site-replication.go:4745-4746`）先 heal Versioning 再 heal Object Lock。我推演过仍收敛（最多 2 个周期、无写入环），但注释与 heal 顺序不一致，建议补一句。
-
---
-
-### F7 — 范围说明（非缺陷，但必须进入决策）
-
-**默认 `MINIO_SITE_REPLICATION_METADATA_TOMBSTONES=off` 时，Tags / SSE / Quota 的"漏发删除"不会通过 heal 收敛。** 只有 Policy 墓碑默认导出（`cmd/site-replication.go:3953-3954` 无条件导出；另外三类在 `:3960 / :3980 / :3988` 被 gate 挡住）。
-
-我完整推演了 off 模式，结论与文档一致：本地真实墓碑在锁内比较时**不会**被旧 PUT 复活（`applyBucketConfig` 用真实落盘状态比较）；但持有旧数据的对端会每 30 秒发一次过期 RPC 被拒绝，**状态永不收敛**，直到运维在权威站点重新提交删除，或全站点升级后统一开启开关。
-
-也就是说：**默认配置下交付的是"顺序正确 + 不复活 + Policy 删除可 heal"，不是"四类删除都能自愈"。** 这是 v4 计划的既定取舍（计划 §提交 3 表格），实现与文档都如实写了；我在此只是确保批准时看到这一点。
-
---
-
-### F8 — 证据不足项
-
-1. **F2 对双站点实验不可见**：`twosite/main.go:141-144` 的所有策略都经 `SRPeerReplicateBucketMeta` → `canonicalBucketPolicy` 写入，两侧都是规范字节，`states()`（`:160-164`）的字节比较自然通过。**从未构造过"升级前旧编码字节"的策略。**
-2. **F1 被 stub 掩盖**：`cmd/site-replication-metadata-gate_test.go:151-159` 自定义 `GetBucketInfo`，在 `opts.NoMetadata` 时直接返回物理 Created，绕过了 `erasureServerPools.GetBucketInfo` 的缓存覆盖。`"physical-created"` 子用例证明的是一个**生产中不会发生**的行为。
-3. **"稳态 0 metadata RPC"的边界**：`twosite/main.go:182-190` 只统计 `event.Bucket == bucket` 的单桶、双站点、**gate=on**。结论有效但窄；gate=off 的稳态不为零（文档已声明），多桶/多站点未观测。
-4. **二进制身份**：`manifest.json` 给了 `silo-final` 的 sha256，但目录里没有 `--version`/build-info 输出（驱动用 `--quiet --json` 启动），`runtime-final.log` 也没有版本行。验收自述"编译信息为 `c8f264f79 + dirty`"是**诚实的**，且日志里的 gate 行为（`gate=on/off`、`off exporter exposed new tombstone`）只能来自第三个提交的生产代码，所以**没有夸大**；但也**不能从证据目录内独立复核**。补一份 `silo-final --version` 或 `go build` 复现即可闭环。
-5. **baseline.log 行号与最终测试文件不一致**（日志 93/105/117 vs 现文件 96/97/109/121）：说明"修复前复现"跑的是测试文件的早期版本。可接受，但严格讲不是同一份用例。
-
---
-
-### F9 — 残留边界：桶世代冲突（计划内已声明，非本轮回归）
-
-各站点 `Created` 不同时，A 的 baseline-live（`at == A.Created`）在 B 上会被重算为 `real`（`newBucketConfigState` 用**目标**的 `created` 判定，`cmd/bucket-metadata-replication.go:215-219`），可能压过 B 的真实墓碑。计划明确排除在收敛承诺外，且 `AddPeerClusters`（`cmd/site-replication.go:458-466`，"only one cluster may have data"）封死了最常见入口。剩余入口：分区期间两站点各自建同名桶，或 #78 接管。基线版本在同场景下是**不确定**的（seed 首个 map 项），所以不算回归。
-
---
-
-### F10 — 覆盖缺口：接管时 `Created` 前移越过真实修改时间
-
-`rebaseBucketConfigDefaults`（`cmd/bucket-metadata-replication.go:306-318`）只调整"零/等于旧 Created"的默认时间。若 `opts.CreatedAt` 晚于某个**真实**字段时间，该字段变成 `at < Created` ⇒ `valid == false`：作为 heal 源 `candidate()` 为 false；作为 heal 目标 `incoming.valid > current.valid` ⇒ **必被覆盖**，本地真实配置被丢弃。
-
-在"远端世代胜出"的语义下可以论证这是对的，但 `TestPeerBucketAdoptionRebasesOnlyDefaults`（`cmd/site-replication-metadata_test.go:394-425`）用的 shift 是 ±1h 而真实时间在 `created+2h`，**恰好没有覆盖这一情形**。建议加一个 `shift = +3h` 的子用例，把期望行为固定下来。
-
---
-
-## 2. 六类配置 × 各入口 覆盖判断
-
-| 配置 | 本地写 | typed peer | bulk | import | initial sync | heal | 接管 | 结论 |
-|---|---|---|---|---|---|---|---|---|
-| Policy | ✅ `bucket-policy-handlers.go:108/152`，hook 用 `result.meta`+`result.updatedAt` | ✅ `site-replication.go:1716-1730` | ✅ `len(item.Policy)!=0` 判"已提供" | ✅ 共同 `commitAt` + 空策略另发专用 nil 事件（`admin-bucket-handlers.go:1150-1155`） | ✅ gate 控墓碑 | ✅ | ✅ rebase | 通过（F2/F4 为附带） |
-| Tags | ✅ `bucket-handlers.go:1940/2021` | ✅ `:1733-1747` | ✅ `*string != nil` | ✅ | ✅ gate | ✅ 补齐 `UpdatedAt`（旧版 heal 缺此字段） | ✅ | 通过 |
-| SSE | ✅ `bucket-encryption-handlers.go:106/199` | ✅ `:1787-1801` | ✅ | ✅ | ✅ gate | ✅ | ✅ | 通过 |
-| Quota | ✅ `admin-bucket-handlers.go:85-99`（删掉零值改写） | ✅ `:2032-2047` | ✅ `len(item.Quota)!=0` | ✅ | ✅ gate | ✅ 含缓存清除（`parse=false` 修复） | ✅ | 通过 |
-| Versioning | ✅ `bucket-versioning-handler.go:100-116`，广播用归一后的 `result.meta` | ✅ 空=no-op | ✅ 空=no-op | ✅ 归一后落盘并广播 | ⛔ 不发（由 MakeBucketHook bootstrap + heal 对齐，与基线一致） | ✅ 按**目标** Lock 状态归一后比较 | ✅ `enablePeerBucketVersioning` 用 `localBucketConfigUpdatedAt` | 通过 |
-| Object Lock | ✅ `bucket-handlers.go:1841-1852` | ✅ 空=no-op，保留 `item.Tags` legacy 回退 | ✅ 空=no-op | ✅ | ✅ | ✅ | ✅ | 通过 |
-
-逐条核对结论：
-
- **来源时间不会变成本地 now**：六个 typed handler 全部删除了锁外 `GetXConfig()` 判旧，统一走 `updateAndParseMetadata(..., &updatedAt)`；非零 `sourceTime` 直接 `sourceTime.UTC()`（`bucket-metadata-sys.go:191-193`）。只有 `sourceTime == nil`（本地写）或为零（legacy 兼容）才分配本地单调时间。✔
- **本地写严格单调**：`localBucketConfigUpdatedAt`（`:255-263`）保证 `> Created` 且 `> 当前字段时间`，涵盖"已有未来时间"。✔
- **重复/乱序不保存不广播**：`applyBucketConfig` 在 `compare <= 0` 返回 `changed=false`，据此跳过 `saveMetadata` 与 `LoadBucketMetadata`；bulk 用 `changed` 汇总后一次保存。✔
-  我特别核对了所有使用 `result.meta`/`result.updatedAt` 的本地 handler：在这些路径上 `changed` 恒为 true（`localBucketConfigUpdatedAt` 使 `incoming.at` 严格更大，且 `incoming.candidate()` 恒真），所以**不会**出现"零时间 + 空载荷"被当成删除广播出去。这是一个隐式依赖，建议加一行注释或断言固定住。
- **整桶 `.metadata.bin` 读-比较-写全在既有分布式锁内**：`updateAndParseMetadata`、`PeerBucketMetadataUpdateHandler`、import 最终提交、`PeerBucketMakeWithVersioningHandler`、CORS 路径，我逐个确认锁的获取在读之前、释放在 `saveMetadata` 之后、fan-out 在释放之后。✔
- **无关字段 / CORS / lifecycle 不被覆盖**：所有写路径都是锁内**重新加载**后只改目标字段；import 用 `applyImportedBucketMetadata` 只按 `fields` 拷贝并 `bytes.Clone`。✔
- **缓存快照正确**：`saveMetadata` 改收 `*BucketMetadata`，`meta.Save` 内部 `parseAllConfigs` 回写归一化结果，`sys.Set(name, *meta)` 发布的是提交后快照；调用方拿到的 `result.meta` 是同一份值拷贝，不会原地修改已发布引用。✔
- **`parse` 被强制为 false 对缓存无害**：`Save()` 先 `parseAllConfigs`，发布到 `metadataMap` 的解析字段是新的；而"删除后 quota 解析残留"正是靠 `parse=false` 的新加载对象修掉的（`parseAllConfigs` 对空 `QuotaConfigJSON` **不会**把 `quotaConfig` 置 nil，`bucket-metadata.go:400-405`）。✔
- **锁/返回值/错误处理**：`unlock()` + `locked=false` 模式保留；`updateAndParse`、`Update`、`Delete` 对外签名不变。✔
-
---
-
-## 3. 六字段 state 比较：传递性、交换序无关、幂等收敛
-
-`compareBucketConfigStates`（`bucket-metadata-replication.go:227-253`）实际是按 `(valid, real, real?at:—, real?isTombstone:—, key)` 的字典序全序。
-
- **传递 / 反对称**：是。`real == false` 时不比较 `at`（只比 key），但空 baseline 的 key 为 nil，`bytes.Compare` 使其恒最小，效果等同"空 baseline 永不获胜"。✔
- **交换顺序无关**：`latestBucketConfig` 取全序最大值，与 map 遍历顺序无关；`applyBucketConfig` 同样只依赖全序。`TestLatestBucketConfigCandidates` 跑了全部 6 种排列。✔
- **幂等收敛**：写入后目标状态等于源状态 ⇒ 下一轮 `compare == 0` ⇒ 不写不发。我另外手工推演了 Object Lock/Versioning 的"Enabled vs Suspended 同时间"场景，归一后双侧都不再写，稳定无振荡。✔
- **baseline Created 与真修改的区分**：`real = valid && at.After(created)`，零字段时间先回填为 `created`（与 `defaultTimestamps()` 一致）。✔
- **同时间删除胜出**：`:244-250`，仅在 `a.real` 分支内生效，空 baseline 不会借此删配置。✔
- **等时不同载荷的稳定键**：
-  - **Policy**：递归排序的规范 JSON + `json.RawMessage` 保留整数精度（测试 `:308-311` 验证 `9007199254740993` 不失真）。我对照了 `silo-pkg .../policy/bucket-policy-statement.go:27-36`，`BPStatement` 的 8 个字段（SID/Effect/Principal/Actions/NotActions/Resources/NotResources/Conditions）**全部覆盖，无字段静默丢失**；`ParseBucketPolicyConfig` 先做 `Validate` 保证 `Principal.MarshalJSON` 不会失败。✔ 但顶层 Statement 排序引出 F2。
-  - **Quota**：`json.Marshal(parseBucketQuota(...))`。零 quota 仍是 live 文档（`candidate()` 靠 `real`，不靠 `len(data)`）。`{}` 与对端重编码字节不同但键相同，不产生 heal 环；`isBktQuotaCfgReplicated` 按解析值比较，不会误报。✔
-  - **Tags / SSE / Object Lock**：有效文档字节，保留大小写与实际内容。✔
-  - **Versioning**：先按**该站点**的 Object Lock 归一（`effectiveBucketVersioning`），键即归一后的文档，避免"保存阶段隐式改写"造成的空转。✔
- **空策略**：`bucketConfigPayload` 对 `cfg.IsEmpty()` 返回 `(nil, nil, nil)`，与 peer 侧既有"空策略=删除"解释统一。✔（副作用见 F4-1）
- **Object Lock 强制 Versioning 后的持久化 vs 广播**：本地只广播 Object Lock 事件，派生的 Versioning 由对端在 `Save → parseAllConfigs` 自行推导，**内容一致、时间各自保留**；heal 用归一后的有效文档比较，最多一次写入即收敛。与计划一致。✔
-
---
-
-## 4. Created / 时间的异常输入
-
-| 输入 | 行为 | 判断 |
-|---|---|---|
-| Created 为零 / metadata.bin 缺失 | 六类写入报错；初次同步静默跳过 | **F1，回归** |
-| 字段时间为零 | 回填为 Created ⇒ baseline | ✔ 与 `defaultTimestamps()` 一致 |
-| 事件时间 == Created | 可更新仍为 baseline 的字段，时间保存为 Created；同时间 nil/空只算空 baseline，不删除 | ✔ 有用例 |
-| 事件时间 < Created | `before-created`，跳过 + 单条日志；heal 对该目标 `continue` | ✔ |
-| 未来时间 | 接受为真实状态；后续本地写用 `max(now, at+1ns)` 压过 | ✔ 有用例 |
-| 各站点 Created 不同 | 见 F9（计划外，已声明） | 残留 |
-| 空 deployment ID | `latestBucketConfig` / `healBucketConfig` 均 `id == "" \|\| !known` 跳过 | ✔ 有用例 |
-| peer 报错 / 缺元数据 | `candidate()` 为 false ⇒ 不参与选源；作为目标 `CreatedAt.IsZero()` ⇒ `continue`。**不会被当成删除** | ✔ |
-| 单个 peer 不可达 | 记录后继续其它目标（`site-replication-metadata.go:182-186`），不因 map 顺序放弃健康站点 | ✔ 有用例（`"broken"`） |
-
-**"已知历史不可恢复"与"新实现回归"的区分**：
- 历史不可恢复（可接受）：旧版到达时间污染、legacy-zero 事件产生的新本地时间、桶世代分歧。
- 新实现回归（应修）：F1 的零 Created 硬失败与初次同步静默漏发；F2 的永久假 mismatch；F3 的 ERROR 噪声与日志互相顶替。
-
---
-
-## 5. 兼容路径与开关
-
- **typed 零时间兼容**：`sourceTime != nil && IsZero()` ⇒ 分配本地时间 + 一条 `legacy-zero`（`bucket-metadata-sys.go:186-190`）。不受 gate 影响，不在源时间排序保证内。✔
- **bulk 零时间拒绝**：`PeerBucketMetadataUpdateHandler:1610-1612` ⇒ `errInvalidArgument`，与现状一致；import hook 恒带非零 `commitAt`。✔
- **省略 vs 显式 null / 空串**：`madmin.SRBucketMeta` 全字段 `omitempty`（`madmin-go v3.0.110 cluster-commands.go:501-538`），因此
-  - `json.RawMessage`（Policy/Quota）：省略 ⇒ nil；显式 `null` ⇒ `[]byte("null")`，走 `len(...)!=0` 判为"已提供"，再交既有解析器（Policy `null` ⇒ 语义空 ⇒ 删除；Quota `null` ⇒ 零值文档 ⇒ live）。✔ 与计划 §1 完全一致。
-  - `*string`：省略/`null` ⇒ nil ⇒ 不动；空串 ⇒ 已提供且内容为空 ⇒ 四类删除、两类 no-op。✔
-  - 这些由 `TestPeerBucketMetadataWireAtomicity` 通过**真实 admin 路由 + 真实 JSON 编解码**覆盖（`applySRBucketMetaViaAdmin` 走 `registerAdminRouter`），不是结构体直传。这是这批测试里质量最高的一块。
- **direct hook 与 heal 的删除传播**：普通 DELETE 事件在 gate off 下照常复制；heal 依赖导出可见性 ⇒ 见 F7。
- **初次同步**：保留原五类范围（不含 Versioning），gate 控制是否发送真实墓碑。✔
- **gate 覆盖点是否遗漏**：我检查了所有导出/初次发送点 —— `SiteReplicationMetaInfo` 的 Tags/Quota/SSE 三处（`:3960/:3980/:3988`）+ `initialBucketConfigReplicationEvent:51`。Policy 墓碑**故意**不受 gate 控制（原本已导出）；Versioning/Object Lock 无墓碑概念。**未发现遗漏的开关覆盖点。** ✔
- **滚动升级/降级**：off 时新增墓碑信息不导出，旧节点不会收到它无法正确处理的 Quota heal 墓碑（文档点名了"旧版会留下已解析 quota 残留"这一实证依据）。✔ 但 off **不等于**与旧版同构：修复端的**接收**行为已经变了（锁内排序、空 baseline 不删除、重复不写），这对旧版发来的事件是更安全的方向，混版冒烟也验证了普通 PUT/DELETE 互通。降级路径文档要求先关开关再滚降；唯一可补的一句是"墓碑时间字段本来就在 schema 内，旧版只是不导出，降级不会产生解析错误"。
-
---
-
-## 6. 三个独立判断
-
-**最小？—— 基本是，有一处可争议的扩张。**
-生产 Go 代码净增约 37 行，六份重复的 heal / peer apply 被一套 helper 替代；没有新 schema、没有新 wire 字段、没有新锁、没有能力协商、没有迁移系统。唯一超出必要的是 **Policy 规范编码器接入写/读/导出路径**（F5）——它不是收敛所必需的，并且直接导致 F2。其余（`bucketMetadataUpdate` 提交快照、`ensureBucketMetadataCreated`、`rebaseBucketConfigDefaults`、gate）我都能各自对应到一个已证实的缺陷或本次修复直接触及的路径，**没有**可以无损删除的部分。
-
-**充分？—— 对"已声明的范围"是；对"缺陷标题"不是。**
-在同一桶世代、Created 已知、gate=on 的前提下，六类配置的源时间、锁范围、删除状态参与 heal、等时冲突裁决、重复/乱序抑制都成立。但两个口子要明说：(1) 默认 gate=off ⇒ Tags/SSE/Quota 的漏发删除不收敛（F7，计划内取舍）；(2) Created 未知的桶从"能写"变成"不能写"（F1，计划外回归）。
-
-**必要？—— 是。**
-每一项改动都能对应到 `baseline.log` 里的实测失败（SOURCE_TIME / NEWER_DELETE / STALE_RESURRECTION / BULK_STALE_OVERWRITE / CHECK_OUTSIDE_LOCK）或计划中论证过的路径。`rebaseBucketConfigDefaults` 这种看起来最"多余"的小分支，实际是 `Created` 一旦参与 baseline/tombstone 判定后的必然补丁（不加则接管时默认值会变成假墓碑）。
-
---
-
-## 7. 新增设计记录应准确保留的内容
-
-**关键决策**
-1. baseline / live / tombstone 三态由 `(Created, 字段时间, 载荷空否)` 推导，**不新增 schema 字段**；`at == Created` 定义为 baseline。代价：必须知道 Created（⇒ F1 的根因）。
-2. 全序：`valid > real > at > 同时间墓碑胜 > 稳定内容键`。deployment ID **不**参与比较，也不落盘。
-3. 比较键是"已解析配置的纯函数"：Policy 走递归排序的规范 JSON，Quota 走解析后 `json.Marshal`，XML 走有效文档字节，Versioning 先按本站点 Object Lock 归一。
-4. Versioning / Object Lock 为 update-only：空事件恒为 no-op，空值不是候选。
-5. 读-比较-写整体在**既有** `metadata.lock` 内；`saveMetadata` 收指针以便调用方拿到提交后快照，保证"落盘状态 / 出站事件 / 源时间"三元组一致。
-6. 新增删除信息的**导出**用启动开关 gate，默认 off；开关不探测远端能力，启用条件是全站点全节点已升级且旧请求排空。
-7. 专用 peer 事件的零时间保留兼容例外（分配本地时间 + 限频 `legacy-zero`）；bulk 零时间仍按现状拒绝。
-
-**被拒绝的方案**（记录以免复议）
- 新增 HLC / 向量时钟 / 新 wire 字段 / 能力协商 / 通用复制框架 / 新锁或重试系统。
- 用 `len(payload)==0 ⇒ 删除` 统一处理 bulk（会把"省略"误判为删除）。
- 把 `BucketPolicy.Equals`（忽略 Sid、对 Statement 顺序敏感）叠加为第二套判等规则。
- 把混版长期测试作为提交门槛。
- 把未修改上游 MinIO 的兼容性当作必需门禁（AGENTS.md：正式支持 PGSTY 栈，上游兼容为尽力而为）。
-
-**限制**
- gate=off 期间 Tags/SSE/Quota 墓碑不可见，不承诺删除收敛，且存在被拒绝的周期性 RPC。
- 旧版到达时间污染、legacy-zero 产生的新本地时间、桶创建世代分歧**无法自动反推**，需运维在权威站点重新提交。
- `Created` 未知的桶不在收敛承诺内（并且按当前实现直接不可写，见 F1）。
- "比较键相同但落盘字节不同"（Quota 原始 JSON、Policy 旧编码）是允许的稳定状态，公开 mismatch 统计必须与比较键口径一致（见 F2）。
-
---
-
-## 8. 建议处置顺序
-
-1. 判定 F1 的条件（支持矩阵里有没有 `Created == 0` / 无 `.metadata.bin` 的桶）。有 ⇒ 先修再合。
-2. 修 F2（推荐：不排序顶层 `Statement`，或让 `isBktPolicyReplicated` 改用规范键）。
-3. 修 F3（ERROR→Warning、缺桶不报、reason 拆分）。
-4. 补 F4 的三条文档。
-5. 补测试：F10 的接管用例、F8-1 的 legacy 策略字节双站点用例、F8-2 去 stub 的 Created 用例、F8-4 的二进制身份记录。
-6. F5 / F6 由作者判断，可留作后续。
@@ -1 +0,0 @@
-ok  	github.com/minio/minio/cmd	6.864s
@@ -1 +0,0 @@
-ok  	github.com/minio/minio/cmd	20.694s
@@ -1,57 +0,0 @@
--- FAIL: TestBucketMetadataPhysicalCreatedRecovery (0.57s)
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/object-lock.xml/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update object-lock.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/versioning.xml/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update versioning.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/policy.json/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update policy.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/tagging.xml/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update tagging.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/bucket-encryption.xml/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update bucket-encryption.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/quota.json/missing=false (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update quota.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/object-lock.xml/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update object-lock.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/versioning.xml/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update versioning.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/policy.json/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update policy.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/tagging.xml/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update tagging.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/bucket-encryption.xml/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update bucket-encryption.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/ErasureSD/quota.json/missing=true (0.00s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update quota.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/object-lock.xml/missing=false (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update object-lock.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/versioning.xml/missing=false (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update versioning.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/policy.json/missing=false (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update policy.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/tagging.xml/missing=false (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update tagging.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/bucket-encryption.xml/missing=false (0.06s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update bucket-encryption.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/quota.json/missing=false (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update quota.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/object-lock.xml/missing=true (0.04s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update object-lock.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/versioning.xml/missing=true (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update versioning.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/policy.json/missing=true (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update policy.json: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/tagging.xml/missing=true (0.01s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update tagging.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/bucket-encryption.xml/missing=true (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update bucket-encryption.xml: bucket metadata creation time is unknown
-    --- FAIL: TestBucketMetadataPhysicalCreatedRecovery/Erasure/quota.json/missing=true (0.02s)
-        site-replication-metadata-gate_test.go:227: bucket without a recorded creation time cannot update quota.json: bucket metadata creation time is unknown
--- FAIL: TestBucketPolicyReplicationStatusLegacyOrder (0.31s)
-    --- FAIL: TestBucketPolicyReplicationStatusLegacyOrder/ErasureSD (0.01s)
-        site-replication-metadata_test.go:348: equivalent legacy and received policy reported as permanently mismatched
-    --- FAIL: TestBucketPolicyReplicationStatusLegacyOrder/Erasure (0.06s)
-        site-replication-metadata_test.go:348: equivalent legacy and received policy reported as permanently mismatched
-FAIL
-FAIL	github.com/minio/minio/cmd	2.940s
-FAIL
@@ -1,8 +0,0 @@
--- FAIL: TestHealBucketConfigDiagnostics (0.29s)
-    --- FAIL: TestHealBucketConfigDiagnostics/ErasureSD (0.00s)
-        site-replication-metadata-heal_test.go:80: empty baselines produced diagnostics
-    --- FAIL: TestHealBucketConfigDiagnostics/Erasure (0.01s)
-        site-replication-metadata-heal_test.go:80: empty baselines produced diagnostics
-FAIL
-FAIL	github.com/minio/minio/cmd	2.721s
-FAIL
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Feng Ruohang	2a4d51406b	test: build the purge-response fixture ARN at runtime The rebrand compatibility guard tracks every literal policy value that names the upstream brand. Construct the replication ARN the way the other replication fixtures do instead of adding a new literal. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 23:50:09 +08:00
Feng Ruohang	0596685ae7	docs: record the delete-marker purge, listing quorum and migration tag repairs Describe the three correctness repairs merged after the 20260916 candidate was prepared, together with their known remaining limits (#217, #218). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 23:44:44 +08:00
Feng Ruohang	358ab38fb0	fix: report a purged version as a delete marker only for stored markers The exact-version purge path copies the looked-up DeleteMarker flag into the DELETE response so that removing a delete marker keeps its x-amz-delete-marker header. That flag is also set for a data version whose purge is pending, because the lookup exposes such a version as deleted for visibility. Retrying the purge after the replication configuration was removed therefore answered with x-amz-delete-marker: true and raised ObjectRemovedDeleteMarkerCreated for a data version. Only a stored marker carries no erasure layout; use that to decide the response identity. Found by the adversarial review of the purge change. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 22:53:12 +08:00
Feng Ruohang	254b19ac07	fix(replication): recheck queued delete-marker creations under the lock A delete-marker creation task carries the marker as it looked when it was queued by the DELETE handler, a GET/HEAD/LIST heal, the scanner or MRF. Tasks from several frontends serialize on the per-object replication lock, so a task can run after the user has purged that marker and the purge has already reached the targets. The target no longer holds the marker, so the queued creation recreated it there with the original VersionID and mtime. This is the late-create sequence seen in the 2026-09-16 three-site runs: purge 204 on both targets, then about 85 ms later a replicated creation for the same VersionID. Re-read the source version under the replication lock before sending a creation. A missing version, a non-marker version, or a version under purge makes the task stale; it is dropped without touching the source. A read that cannot confirm either way is retried through MRF instead of being treated as absence. Creations already on the wire, replays from other sites, and cleanup of minority residue after a crash are not covered by this check. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 22:35:08 +08:00
Feng Ruohang	eb4f5e5b31	Fix exact-version purges and delete-marker metadata healing Require write-quorum absence for missing purge retries, aggregate removed and already-absent votes only for physical purges, and resolve receiver purges by version across pools. Preserve marker metadata through creation and healing while retaining DELETE response semantics. Add real-disk quorum, pool, callback, metadata, heal and outbound replication regressions. Synchronize capacity fixture installation and restoration with concurrent IAM readers. Signed-off-by: Feng Ruohang <rh@vonng.com> (cherry picked from commit 22ba4d426677aa07470927fc349a43afd87e19b6)	2026-09-16 22:24:27 +08:00
Feng Ruohang	8d06424b12	fix: retain quorate null versions after newer minorities Recount only when the existing selection is below quorum and each original nonempty stream has one ordinary null version with identical EC headers. Reuse the existing header grouping and leave history pruning unchanged. Add deterministic signed LIST, complete-metadata permutation, excluded-history, object-read, scanner/heal, Walk and migration-consumer regressions. Signed-off-by: Feng Ruohang <rh@vonng.com> (cherry picked from commit 9804e4deb4d5a18cca640be72bff9b47a415f7de) Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 21:50:03 +08:00
Feng Ruohang	fced863036	fix(storage): preserve tags during pool migration Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 20:22:37 +08:00
Feng Ruohang	5af0865aab	Merge pull request #214 from pgsty/codex/release-prep-20260916 Prepare the SILO 20260916000000 source candidate with released component pins, CPU metrics synchronization, regression coverage and release tooling updates. All 11 checks passed on `e7963381ba`. Signed release artifacts, image publication and deployment remain separate deliverables tracked by #203. Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 19:30:07 +08:00
Feng Ruohang	83821f0f1f	Merge pull request #216 from pgsty/codex/docs-published-design-links-20260916 docs: link retained guides to published design records	2026-09-16 18:50:28 +08:00
Feng Ruohang	bf053386a4	docs: link compression and federation to maintained design records Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 18:26:03 +08:00
Feng Ruohang	e7963381ba	test: synchronize tag replication capacity fixtures Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 17:37:47 +08:00
Feng Ruohang	37c0edc7ca	test: use canonical octal permissions in LZ4 fixture Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 17:33:58 +08:00
Feng Ruohang	a2fe70424e	release: prepare SILO 20260916000000 candidate Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 17:30:29 +08:00
Feng Ruohang	027d43b4eb	fix: synchronize CPU resource metrics reads Hold resourceMetricsMapMu.RLock while the v3 CPU collector reads both the subsystem map and its nested idle/iowait metrics. The unsynchronized reads were inherited from upstream commit `f7b665347` (minio/minio#19560). Add value and concurrent Prometheus Gather regression tests, run them in the targeted race CI job, and record the existing CPU metric names in the compatibility inventory. Fixes #210 Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 17:09:19 +08:00
Feng Ruohang	f99ed829b5	Merge pull request #213 from pgsty/codex/pr198-release-compat fix: preserve multipart defaults and rollback compatibility	2026-09-16 16:42:56 +08:00
Feng Ruohang	956a1a8e33	Merge pull request #212 from pgsty/codex/docs-migration-cleanup docs: consolidate repository entry points and retire migrated work records	2026-09-16 16:39:40 +08:00
Feng Ruohang	82f0a9828e	fix: preserve multipart defaults and rollback compatibility Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 16:30:43 +08:00
Feng Ruohang	584b5a3a8f	docs: link the canonical recovery runbooks Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 16:19:56 +08:00
Feng Ruohang	39f49d548f	Merge remote-tracking branch 'origin/main' into codex/docs-migration-cleanup Signed-off-by: Feng Ruohang <rh@vonng.com> # Conflicts: # CONTRIBUTORS.md	2026-09-16 16:13:31 +08:00
Feng Ruohang	a168576adb	Merge pull request #208 from pgsty/codex/conditional-put-release-notes docs: describe conditional PUT behavior and tested recovery guidance	2026-09-16 16:09:22 +08:00
Feng Ruohang	e85c4c8dfb	docs: preserve reviewed contributor records and daily card references Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 16:09:00 +08:00
Feng Ruohang	0e3c43778e	Merge pull request #211 from pgsty/codex/daily-repository-cards ci: update README repository cards daily at 00:00 UTC	2026-09-16 15:40:09 +08:00
Feng Ruohang	9101fe78db	ci: refresh README repository cards daily at 00:00 UTC Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 15:27:48 +08:00
Feng Ruohang	82948d6306	Merge pull request #198 from mrjavadseydi/fix/issue-79-list-multipart-uploads Retain the original contributor commit and integrate the reviewed durable listing, cancellation confirmation and migration fixes. Capacity acceptance and delayed creation-write fencing remain open in issue #79. Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 13:40:10 +08:00
Feng Ruohang	dfb4b2a1d2	fix: return HTTP 503 when multipart scan capacity is exhausted Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 13:24:52 +08:00
Feng Ruohang	143f6970d8	fix: make multipart discovery and cancellation explicit and bounded Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 13:16:37 +08:00
Feng Ruohang	ea5dac5a99	Merge current main into multipart listing contribution Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 12:51:20 +08:00
Feng Ruohang	3c26a8b0b5	Merge pull request #209 from pgsty/codex/credit-console-share-reporter fix: integrate the bounded Console sharing proxy and credit its reporter	2026-09-16 12:19:29 +08:00
Feng Ruohang	2fabd436c0	fix: select the bounded Console sharing proxy Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 12:05:00 +08:00
Feng Ruohang	07c68d6054	docs: credit Jiri Pejchal for the Console sharing report Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 11:48:38 +08:00
Feng Ruohang	15090dc4fd	docs: retire migrated working material and clarify documentation ownership Remove migrated investigation and security documents, repair their incoming links, and align the English/Chinese READMEs with current module and release boundaries. Track AGENTS.md as the shared repository guide and make CLAUDE.md import it; keep working artifacts outside the repository. Publish pgsty/silo.pgsty.com commit c7682185e2832b981629e1c17aef74fc778c6ca1 before publishing this cleanup: the new documentation routes are not live yet. Validation: make rebrand-guard; 92 Markdown files checked for deletion-induced broken relative paths; 199 source-to-site references and anchors resolve in the paired site build; git diff --check. Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 11:22:48 +08:00
Feng Ruohang	e791640dac	docs: describe merged conditional PUT behavior and recovery guidance Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 11:21:16 +08:00
Feng Ruohang	9b4ae82a29	Merge pull request #207 from pgsty/codex/conditional-put-pools fix(pools): evaluate cross-pool PUT conditions against the current object	2026-09-16 11:17:02 +08:00
Feng Ruohang	4620be394b	test: synchronize conditional PUT disk fixtures Replace unsynchronized getDisks swaps with backing disk-list updates under erasureDisksMu, matching the existing GetDisks reader lock. Apply the same helper to capacity and read-fault adapters while preserving nested restore ordering. Add a regression that overlaps fixture changes with the real IAM Walk reader, and run the conditional PUT suite under the race detector in CI. The regression reproduces the old fixture race; ten fixed race iterations pass without warnings. Production conditional PUT behavior is unchanged. Refs #199 Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 10:53:24 +08:00
Feng Ruohang	5e7d603083	fix(pools): evaluate cross-pool PUT conditions against the current object Multi-pool PUT selected a destination by capacity and evaluated If-Match/If-None-Match only against that destination's local object state. An empty or stale destination could accept a stale ETag or If-None-Match:* while another pool held the current object, replacing it; a current ETag could instead be rejected with 412 or 404. Under PUT's existing pools-layer object lock, resolve the comparison object with objectPoolInfos (including draining pools), treat a latest delete marker as absence, fail closed on unreadable pool metadata, and clear an accepted callback before destination dispatch. Replica and data-movement callbacks keep their addressed-version semantics and metadata reconciliation. Reproduced on `40220bd836` and RELEASE.2026-09-03T13-18-01Z with six signed HTTP scenarios: four defect cases failed, two controls passed. Refs #199 Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 08:21:21 +08:00
Feng Ruohang	fb7c406ddc	Merge pull request #206 from pgsty/codex/main-consolidation-20260916 test: restore valid credentials in integration fixtures	2026-09-16 08:18:56 +08:00
Feng Ruohang	416826f61f	Merge pull request #205 from pgsty/codex/release-notes-followups docs: complete September correctness and upgrade notes	2026-09-16 08:16:23 +08:00
Feng Ruohang	a2e2f3ee82	test: restore valid credentials in integration fixtures Port the shell-fixture changes from ebc9937d97b27871dcc4bb91d4b5771d3550b76a. Match the existing minimum secret length consistently across startup, aliases and helper commands. Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 08:08:08 +08:00
Feng Ruohang	70c7ec4a9f	docs: link reviewed runbook sources before website deployment Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 08:04:53 +08:00
Feng Ruohang	2b722b7e92	docs: complete September correctness and upgrade notes Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 07:51:11 +08:00
mr javad seydi	4cbb074ccd	fix: make multipart upload listing S3-compatible Signed-off-by: mr javad seydi <seydi.birjand@gmail.com>	2026-09-15 23:17:55 +03:30
Feng Ruohang	40220bd836	Merge pull request #196 from pgsty/codex/merge-r4-r8 Complete R5, R6 and R8 on the main baseline containing R4 and R7. Preserve the individual signed repairs, independent Opus 5 Max review, and integration validation evidence. Signed-off-by: Feng Ruohang <rh@vonng.com>	2026-09-16 01:14:06 +08:00