1.39.6 — Project health checks and fleet readiness

2026-10-22

Optional project service checks join existing health snapshots, and fleet monitoring uses current signed request readiness.

Added

• Optional trusted project service checks in the existing five-minute health snapshots, with deadlines, permissions and availability evidence.
• Configuration fingerprints invalidate cached project health when check declarations or routing expectations change.

Fixed

• Fleet node diagnostics, monitoring coverage and snapshots now share current signed request-process readiness.
• Proved readiness failures remain failures; stale, missing or mismatched proof remains Unknown. Cached polling cannot renew observations.

Upgrade notes

• Project checks remain disabled until configured through HEALTH_SNAPSHOT_PROJECT_SERVICES. Callbacks must use bounded I/O and include cleanup within their deadline.
• Consumers of configured project checks must support the new service component kind or reject unsupported snapshots safely.
• Update job workers alongside API processes to collect current fleet readiness evidence.