Recover — Stop the Bleeding, Then Diagnose

Get production healthy fast and safely, in the right order: recover first, root-cause second. A live incident is not the time to debug in prod or to fix forward a change you don't understand yet — revert to the last known-good, then investigate calmly.

North star: shortest safe path back to known-good, a clear status thread while it's open, and a written follow-up so it doesn't recur — without a panicked command making it worse.

Step 0 — Triage (60 seconds)

• What's the blast radius? One endpoint, one tenant, or everything? How many users affected?
• What changed most recently? The last deploy is the prime suspect — check deploy history, git log on the release branch, and recent config/flag changes.
• Open a status thread now in your incident channel: one line — what's wrong, who's on it, "investigating." Update it, don't spawn new messages.

Step 1 — Stop the bleeding (recover before you diagnose)