v5.3.1 — notify when a second health check fails

Patch release. Doctor health alerts now fire when a new check starts failing even if another equally severe check is already failing.

npm @canonry/canonry@latest is 5.3.1.

Fixes (from #1167)

• Notify on the failing set, not only the headline code. Health alerts used to fire only when the worst check's (status, code) changed. Equally severe checks are ranked by id, so a breach opening under an alphabetically earlier one stayed silent: it was graded, listed in the payload, and then dropped at the trigger. doctor_health_state now records a signature of every failing check (sorted status:code pairs). The trigger fires when that set moves. Rows from before this column read as unknown rather than changed, so deploying this does not page every already-degraded project.
• gbp.data.recent-sync warns at 4 days instead of 7. GBP metrics land daily, so a week of silence was already a long outage.

Also in 5.3.0 (npm, no prior GitHub release)

• ga.data.recent-data / gsc.data.recent-data — alert when daily GA4 or Search Console data stops arriving (#1165).
• site.reachability — alert when the project website stops answering, via a seeded 10-minute site-liveness schedule.

Notes

• No API break. Existing webhooks still receive health.degraded / health.recovered.
• CLI/MCP already tell agents when a newer canonry is published (UPDATE_AVAILABLE on stderr, MCP initialize / canonry_help, canonry.version.current).

Install

npm install -g @canonry/canonry@5.3.1
canonry --version

Then restart any running canonry serve.