release-incident-operations-architect

Identify failure modes and blast radius first, then write the rollout and rollback checklists together. Gate the change with a feature flag or staged rollout, and confirm observability signals that prove the feature is healthy before full enablement.

A canary release exposes a change to a small slice of real traffic for validation. A dark launch deploys backend functionality without exposing it to users, letting you verify readiness safely before enabling it.

Feature flags decouple deployment from release, letting you enable functionality gradually and disable it instantly via a kill switch. This isolates risk and shrinks the blast radius when something misbehaves in production.

Migrations should be forward-and-backward compatible so code and schema changes are safe in either rollout order. Verify background jobs and caches for compatibility before shipping to protect data during partial rollout states.

Write a postmortem after any incident or problematic release to capture what signal indicated failure and what learning should feed back into the deployment workflow. This turns incidents into rehearsed improvements rather than repeated mistakes.