fix(review): forced re-run steals the AI-review lock; re-run checkbox ignores closed/merged PRs

[JSONbored]

Summary

Two related fixes to the panel re-run mechanism:

#9008 — forced re-run silently died behind an orphaned AI-review lock. claimAiReviewLock had no bypass, so a maintainer's forced re-run (the panel checkbox / @loopover review) landed on the exact same contended-placeholder path a genuine race would — with zero audit trail, since every ai_review_* event lives inside the lock-acquired branch. Worse, the placeholder's plain-prose notes masqueraded as a real review summary, suppressing the one alarm (ai_review_public_summary_missing) meant to catch "no fresh review happened." Confirmed live: an orphaned lock (from a self-host process that died mid-review) made the re-run button inert for 15+ minutes.

Fix:

• New steal option on claimTransientLock (cache path: unconditional set instead of claim; SubmissionLock DO: same overwrite semantics), threaded through claimAiReviewLock and used whenever forceAiReview: true — a duplicate LLM call is the explicitly accepted cost.
• The contended path now always names itself in the audit trail (github_app.ai_review_lock_contended, with a forced flag), forced or not.
• Fixed the masquerade by checking persistable === false directly in the alarm condition — the one signal aiReviewLockContendedResult uniquely produces.
• Threaded forceAiReview into shouldStartAiReviewForAdvisory to bypass ONLY the reputation-based skip, never a maintainer-configured hard gate (aiReviewMode off, an ineligible author, etc.).

#9020 — the re-run checkbox does real work (and worse) on an already-merged PR. The panel comment is deliberately preserved after merge (#preserve-review-on-close) and its checkbox stays interactive on GitHub forever, but maybeProcessPrPanelRetrigger never checked pr.state, so a post-merge click ran deep into the pipeline — real GitHub/DB spend — before finally dying at the freshness guard with contradictory telemetry and no user-visible feedback. Second harm: it could reach createOrUpdateSkippedGateCheckRun, which had no updateExisting mode and could PATCH an already-completed, fully-evaluated Gate run down to skipped — falsifying the historical verdict on a merged PR.

Fix: added the same pr.state !== "open" guard reReviewStoredPullRequest already has (mirrors its sibling entry point exactly), and scoped createOrUpdateSkippedGateCheckRun to updateExisting: "in_progress_only" — matching its actual purpose (finalize an in-progress evaluation that never finished) without ever touching an already-terminal run.

Closes #9008, #9020

Test plan

• 4 new transient-locks.test.ts tests: steal on both the cache and DO paths, a genuine (non-steal) claim still contends normally, steal fails open when the cache errors
• ai-review-advisory.test.ts: forceAiReview bypasses reputation but not a hard gate
• queue.test.ts: two full-pipeline integration tests — forced steal + audit event + masquerade fix end to end, plus the audit write's own fail-open branch
• queue-4.test.ts: a re-run click on a merged PR is skipped immediately with pr_not_open, no live GitHub calls
• github-app.test.ts: an already-completed Gate check run is never demoted, a separate run is posted instead
• 100% line+branch coverage on every changed line across all 5 touched source files (verified via lcov)
• npx tsc --noEmit, check-migrations.ts, check-schema-drift.ts, check-engine-parity.ts all green
• Full regression pass across all 11 affected test files (1206/1206 passing)

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.66%. Comparing base (f2f32cc) to head (6a8bd8d).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9108      +/-   ##
==========================================
- Coverage   93.88%   92.66%   -1.23%     
==========================================
  Files         813      813              
  Lines       80666    80677      +11     
  Branches    24477    24479       +2     
==========================================
- Hits        75737    74759     -978     
- Misses       3562     4842    +1280     
+ Partials     1367     1076     -291     

Flag	Coverage Δ
backend	93.47% <100.00%> (-1.70%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/github/app.ts	97.94% <ø> (ø)
src/queue/ai-review-orchestration.ts	100.00% <100.00%> (ø)
src/queue/processors.ts	95.62% <100.00%> (+<0.01%)	⬆️
src/queue/submission-lock.ts	100.00% <100.00%> (ø)
src/queue/transient-locks.ts	100.00% <100.00%> (ø)

... and 3 files with indirect coverage changes

[loopover-orb]

Tip

✅ LoopOver review result - approve/merge recommended

_{Review updated: 2026-07-26 19:09:36 UTC}

10 files · 1 AI reviewer · no blockers · readiness 88/100 · CI green · clean

✅ Suggested Action - Approve/Merge

• safe to merge

Review summary
This PR closes #9008 (forced re-run steals an orphaned AI-review lock and now always names itself in the audit trail, plus fixes the lock-contended masquerade via `persistable === false`) and #9020 (panel re-run checkbox now checks `pr.state !== "open"`, and `createOrUpdateSkippedGateCheckRun` is hardened to `updateExisting: "in_progress_only"` so it can never demote an already-completed Gate run). The `steal` option correctly does an unconditional overwrite (`set`/DO overwrite) instead of a CAS `claim`, and `forceAiReview` is threaded through `shouldStartAiReviewForAdvisory` to bypass only the reputation skip, not hard gates — verified in the diff and tests. Tests are substantial and appear to exercise the real paths (orphan-lock steal, contended audit trail, persistable-false alarm, merged-PR panel skip, completed-check-run non-demotion).

Nits — 6 non-blocking

• src/queue/transient-locks.ts: the steal branch's catch block returns `{ acquired: true, ownerToken: null }` on a `set()` failure — this fails open for a forced re-run, which is consistent with existing claim-failure posture, but worth a one-line note that a failed steal means the caller can't ever release the lock it thinks it claimed (transient-locks.ts).
• src/queue/ai-review-orchestration.ts: `aiReviewLockContendedResult` sets `persistable: false` but the new gate check `aiReview?.persistable === false` relies on no other producer ever setting that field false — worth a lightweight type-level guarantee (e.g. a shared constant/type) rather than relying on convention across files.
• src/github/app.ts: the `updateExisting: "in_progress_only"` change to `createOrUpdateSkippedGateCheckRun` silently changes behavior for any other existing caller relying on the previous default 'any' semantics — confirm no other call site depended on demoting a completed run to skipped.
• Consider extracting a small enum/type for the aiReview result shapes (`persistable: false` vs `cacheable: false`) so future producers can't accidentally omit setting `persistable` and silently defeat the alarm again.
• Double check all call sites of `createOrUpdateSkippedGateCheckRun` to confirm none intentionally relied on the old 'demote a completed run' behavior.
• PR author also opened the linked issue — Link an issue that was opened by a different contributor, or provide a rationale for why this self-authored issue represents genuine discovery work.

Decision drivers

• ✅ Code review — No blockers (1 reviewer)
• ✅ Gate result — Passing (No configured blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	✅ Linked	#9008
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	❌ 8/20	High review scope from cached public metadata (1 linked issue).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 13 registered-repo PR(s), 13 merged, 315 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 13 PR(s), 315 issue(s).
Improvement	✅ Minor	risk: clean · value: minor

Linked issue satisfaction

Addressed
The PR wires forceAiReview through to claimAiReviewLock with a new steal option that unconditionally overwrites the lock (both cache and DO paths), adds an unconditional audit event on the contended path, fixes the masquerade by checking persistable === false in the missing-summary alarm, and threads forceAiReview into shouldStartAiReviewForAdvisory to bypass only the reputation skip — covering al

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Python, TypeScript, Ruby, Go, MDX, Shell, Solidity, JavaScript
• Official Gittensor activity: 13 PR(s), 315 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Start here: Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Then work through the remaining 2 steps in the Signals table above.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

Decision record

• action: merge · clause: success
• config: 03a7f8b529a9 · pack: oss-anti-slop
• record: 33a64a18679d (schema v3, head 6a8bd8d)

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review

[loopover-orb]

LoopOver approves — the gate is satisfied and CI is green.