A deployment incident has a documented rollback trigger, but engineers debate ad hoc actions while impact grows. Which operational preparation would most improve response consistency?

1. Maintain tested runbooks with decision criteria, accountable roles, escalation, and recovery verification.
2. Rely only on whoever happens to remember the last incident.
3. Treat incident closure as automatic once a rollback command is typed.
4. Delete all rollback criteria so every responder can improvise freely.