[SSoggyTacoMan]

Summary

When a maintainer updates a PR branch with main or an empty commit is pushed without touching submission game files:

1. Review state was lost or bypassed: In review-state.mjs, CHANGES_REQUESTED and APPROVED strictly matched r.commit_id === headSha. A merge commit changed headSha, which falsely cleared CHANGES_REQUESTED back to Ready for Playtest (as seen on [Sprig App] Galaxy shooter #4046) or demoted an APPROVED PR back to Ready for Playtest.
2. Stale timer corruption: In stale-review-queue.mjs, author commits did not count towards author activity (only issue comments did), and handleUntouchedSubmission used pullRequest.updated_at (which maintainer branch updates refreshed).

Changes

• Diff-aware review reconciliation: Added isReviewCommitOutdated in review-utils.mjs comparing the diff between review.commit_id and headSha for submission files in games/.
• Review state preservation: Updated reconcileReviewStatus in review-state.mjs to accept an outdated check, preserving changes_requested and approved states across commits that do not touch game files.
• Triage and review sync integration: Wired auto-triage.mjs and sync-reviews.mjs into isReviewCommitOutdated.
• Author activity tracking: Updated stale-review-queue.mjs to count author commits toward activity and ignore automated/maintainer updates when computing untouched submission age.
• Unit tests: Added unit tests verifying review status preservation across non-game commits and isReviewCommitOutdated compare diff checking.

[vercel]

@SSoggyTacoMan is attempting to deploy a commit to the Hack Club Team on Vercel.

A member of the Team first needs to authorize it.

[LucasHT22]

There’s an edge case with cherry picked commits but lgtm for now