fix(github): self-heal the reviews cache on the periodic sweep, not just webhooks

[JSONbored]

Summary

• The durable reviews cache (feat(github): extend the head-SHA snapshot cache to live PR reads and reviews #2537, merged in feat(github): cache PR reviews with webhook-based invalidation #2633) is invalidated only by an explicit write on a pull_request_review/synchronize webhook. If that write is ever dropped, nothing else re-checks it — the periodic re-gate sweep only calls refreshPullRequestDetails when slop evidence, the manifest gate, or a pre-merge check path is active, so a "quiet" PR with none of those enabled could carry a stale reviews cache indefinitely.
• Extracts the reviewsUpToDate check out of fetchAndStorePullRequestDetails into an exported isReviewsCacheUpToDate predicate (same reviewsSyncedAt vs reviewsInvalidatedAt comparison, now one authoritative definition instead of a duplicated copy).
• The per-PR sweep unit (reReviewStoredPullRequest) now checks that predicate directly and forces a refresh when the cache is stale and no other reason already triggers one. A failed cache-state read fails open toward "stale" rather than crashing the sweep.
• Tradeoff: a failed sync-state read is intentionally treated as stale (fail-open toward re-checking), and the resulting refresh call is itself still best-effort via refreshPullRequestDetails(...).catch(() => undefined) — a D1 hiccup never crashes the sweep, but it also means that specific pass can be a no-op; convergence relies on a later sweep pass once the read succeeds.

Scope

• src/github/backfill.ts, src/queue/processors.ts — no schema/API/wrangler changes, no migration needed (reuses the existing reviewsInvalidatedAt/reviewsSyncedAt columns from feat(github): cache PR reviews with webhook-based invalidation #2633).

Validation

• npm run test:ci (full local gate: actionlint, migrations check, typecheck, unsharded coverage, workers, MCP build/pack, miner build, REES tests, OpenAPI checks, UI lint/typecheck/test/build) — green.
• npm audit --audit-level=moderate — 0 vulnerabilities.
• New regression tests: sweep forces a refresh on a stale reviews cache with no other trigger active; a failed sync-state read fails open without crashing the sweep. Both mutation-verified (fail without the fix).
• 100% branch coverage on every new/changed line (confirmed directly against the coverage report, not just the aggregate summary).

Safety

• No secrets, tokens, or private scoring values touched.
• No auth/CORS surface touched.

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ Gittensory review result - manual review recommended

_{Review updated: 2026-07-03 00:46:56 UTC}

4 files · 1 AI reviewer · no blockers · readiness 93/100 · CI green · clean

⏸️ Suggested Action - Manual Review

• Touches a guarded path — held for manual review

Review summary
The change moves the review-cache freshness decision into a shared predicate and wires the periodic per-PR sweep to refresh details when that predicate says the stored review state is stale, including a bounded-age fallback for missed invalidation writes. The main path is coherent: existing invalidation ordering still fails closed, failed sweep-side sync-state reads are treated as stale, and the new tests cover stale markers, missing markers past the age window, and malformed timestamps. The notable tradeoff is that cache freshness is now time-dependent in a helper described as pure, so tests and callers need to keep controlling time where cache-hit behavior matters.

Nits — 6 non-blocking

• nit: src/github/backfill.ts:1984 describes isReviewsCacheUpToDate as a pure predicate, but the function reads Date.now(), so either pass now in explicitly or update the comment to avoid misleading future callers.
• nit: src/github/backfill.ts:1976 hard-codes the 48h backstop without an exported name or caller-supplied clock, which makes boundary tests and future tuning more awkward than necessary.
• nit: test/unit/queue.test.ts:1077 adds three large near-identical fetch stubs for the sweep regression cases; a small local helper would make the scenario differences easier to audit.
• In src/github/backfill.ts:isReviewsCacheUpToDate, consider accepting an optional nowMs/now parameter so production can default to Date.now() while tests can assert the 48h boundary without global fake timers.
• In test/unit/queue.test.ts, extract the repeated PR seed and GitHub fetch stub setup from the new regression tests so each test highlights only the cache-state condition it is proving.
• Touches a guarded path — held for manual review — A maintainer must review and merge this change.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	⚠️ Missing	No linked issue or no-issue rationale found.
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (size label size:M; no linked issue context).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 61 registered-repo PR(s), 52 merged, 506 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 61 PR(s), 506 issue(s).
Gate result	⚠️ Not blocking	Advisory; not blocking this PR.

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository registration is not available in the local Gittensory cache.
• Public profile languages: not available
• Official Gittensor activity: 61 PR(s), 506 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Explain no-issue PR.
• No action.
• Link the issue being solved, or explicitly explain why this is a no-issue PR.

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.

_{🟩 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 Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.09%. Comparing base (0658c41) to head (1403887).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2639   +/-   ##
=======================================
  Coverage   96.09%   96.09%           
=======================================
  Files         237      237           
  Lines       26472    26482   +10     
  Branches     9602     9606    +4     
=======================================
+ Hits        25438    25449   +11     
+ Misses        425      424    -1     
  Partials      609      609           

Files with missing lines	Coverage Δ
src/github/backfill.ts	96.96% <100.00%> (+0.02%)	⬆️
src/queue/processors.ts	92.45% <100.00%> (+0.05%)	⬆️

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.