fix(review): skip re-publishing a PR surface already current at its head (#1615)

[JSONbored]

Summary

The scheduled re-gate sweep re-reviewed + re-published every open PR's public surface every ~2-min cycle regardless of change (~4292 pr_public_surface_published in 2h for ~64 PRs). The comment was PATCHed unconditionally and there was no per-PR record of the head the surface was last published at, so the sweep couldn't tell a current PR from a changed one. (AI spend was already deduped via the head_sha-keyed ai_review_cache; this was GitHub-write/processing churn.)

This adds a head_sha-keyed published marker and a sweep skip-guard so the engine only re-reviews/re-publishes non-reviewed or changed PRs:

• pull_requests.last_published_surface_sha (migration 0080) — stamped (head_sha-pinned) after a successful publish. Omitted from the GitHub-sync SET clause like the sibling approved_head_sha/merge_blocked_sha markers, so a later sync can't clobber it.
• Skip-guard in reReviewStoredPullRequest, placed after the live-head resync (so the compare uses the live head): skip the re-review + re-publish when last_published_surface_sha === head_sha. A never-published PR (NULL marker) or a changed head (push/rebase/force-push → marker ≠ live head) falls through and re-reviews at the new head. The webhook synchronize/opened paths review directly and always re-stamp — this guard only gates the sweep.
• Comment idempotency — skip the sticky-comment PATCH when the rendered body is byte-identical (defense-in-depth; also collapses duplicate webhook deliveries).
• Flake fix (folded in, no separate PR) — proof-of-power-stats.test.tsx count-up test forced the prefers-reduced-motion path so the value settles synchronously instead of racing a 3s findByText on the rAF tween (it flaked the ui job on fix(selfhost): route single-provider AI through the name-aware router (#1610) #1611 and fix(selfhost): surface the CLI's structured stdout error before the exit code (#1612) #1613). Verified deterministic across 5 runs.

Rebase re-review is explicitly preserved and pinned by a test: a rebased PR's old-head marker ≠ the live head → it resyncs + re-reviews at the new head (the head_sha-keyed AI cache also misses) → it's never skipped. Expected effect: pr_public_surface_published drops from ~1/PR/sweep to ~1/PR/head_sha.

Closes #1615. Also resolves the #7 ui flake (no separate PR).

Scope

• The PR title follows type(scope): short summary Conventional Commit format.
• This PR is focused (review-engine over-publish dedup + rebase preservation; the only non-backend file is the flake-fix test, folded per maintainer request — no separate PR).
• This follows CONTRIBUTING.md and does not reintroduce GitHub Pages, VitePress, site/, or CNAME.
• I linked an issue.

Validation

• git diff --check
• npm run actionlint
• npm run typecheck
• npm run test:coverage locally; codecov/patch ≥97% — covered: skip-guard (skip / rebase-fall-through / marker-null), markPullRequestSurfacePublished (stamp / null no-op / stale-head no-op), the publish stamp call + its fail-open catch, and both html_url arms of the comment-idempotency early return.
• npm run test:workers
• npm run build:mcp
• npm run test:mcp-pack
• npm run ui:openapi:check
• npm run ui:lint
• npm run ui:typecheck
• npm run ui:build
• npm audit --audit-level=moderate
• New branches + a regression test for the bug; the flake fix verified deterministic (5 runs)

If any required check was skipped, explain why:

• No OpenAPI/cf-typegen regeneration: no route/schema or wrangler.jsonc change. The DB change is a hand-written contiguous migration 0080 (the repo's convention for pull_requests columns — see 0053_pr_approved_head_sha.sql, 0062_pr_last_regated_at.sql); db:migrations:check passes.

Safety

• No secrets, wallets, hotkeys, trust scores, private rankings, or maintainer evidence exposed — the marker is a commit SHA; the skip log carries repo/PR/headSha only.
• Public GitHub text stays sanitized, low-noise — this strictly reduces GitHub writes.
• Auth, cookie, CORS, GitHub App, Cloudflare, or session changes include negative-path tests — N/A.
• API/OpenAPI/MCP behavior updated/tested where needed — N/A.
• UI changes use live API data / real states — N/A (the only UI change is a test-determinism fix).
• Visible UI changes include UI Evidence — N/A (no visible UI change).
• Public docs/changelogs updated where needed — N/A.

Notes

• The marker is intentionally head_sha-keyed (not a content digest): a rebase is the case that MUST re-review, and head_sha is the single freshness key already used by the AI cache and the sibling markers — avoiding a second source of truth.

[superagent-security]

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

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	gittensory-ui	5e8d319	Commit Preview URL Branch Preview URL	Jun 27 2026, 07:06 PM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.51%. Comparing base (050e751) to head (5e8d319).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1616   +/-   ##
=======================================
  Coverage   95.51%   95.51%           
=======================================
  Files         204      204           
  Lines       22041    22050    +9     
  Branches     7963     7966    +3     
=======================================
+ Hits        21052    21061    +9     
  Misses        413      413           
  Partials      576      576           

Files with missing lines	Coverage Δ
src/db/repositories.ts	96.14% <100.00%> (+<0.01%)	⬆️
src/db/schema.ts	68.32% <ø> (ø)
src/github/comments.ts	96.77% <100.00%> (+0.10%)	⬆️
src/queue/processors.ts	88.29% <100.00%> (+0.03%)	⬆️

🚀 New features to boost your workflow:

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