Runbook: Bad Deployment / Rollback

Project-Agent-trust-merge operations docs/ops/runbooks/bad-deployment-rollback.md

Severity: SEV-1 when a new deploy causes customer-visible outage; otherwise SEV-2.
Page: release owner and on-call engineer.
Triage:
Check /health/components and API/frontend health gates.
Inspect deployment-run blockers and latest rollout notes.
Compare pre/post-deploy error-rate and latency deltas.
Mitigation:
Pause rollout activity immediately.
Roll back the most recent deploy or deployment wave.
Keep live cohort rollout disabled until SLOs stabilize.
Customer comms:
Status page update immediately for customer-visible regression.
Follow with direct customer notice if approval, audit, or SSO paths were impacted.