The forge says on the board, and pages once, when production has fallen behind main's newest release for more than an hour #2612

Why: production served 0.7.162 from 2026-10-16 16:34Z while main recorded 0.7.163, 0.7.164 and 0.7.165, plus 161 commits. Nothing on the board said so. Every fix in that window (the restore drill, off-box backups that fit the uplink, the sentinel's no-refile rule) stayed on main while the forge kept reporting the faults they fix. #2574 proves a release after it boots, so it cannot see a release that never boots.

Done looks like:

• [ ] Every 10 minutes, the forge compares its own version (the one /healthz sends as X-Forge-Version) with the newest release in src/core/releases.json on the default branch of its own repository, read from the git store, not the network.
• [ ] When the newest release has been on main for over an hour and is not the version serving, a release vitals cause says 'production is <live>; main released <newest> <age> ago'. The board's forge strip shows it, and the existing incident pager pages once. The cause clears when they match.
• [ ] Tests with an injected clock and release list cover: matching versions say nothing; behind by 30 minutes says nothing; behind by 2 hours raises the cause once; catching up clears it.

Files: a new src/release/behind.ts, the vitals cause, tests. Do not touch .forgejo/ or charts/.