The runbook is part of the design

A failover plan that only works when one particular engineer runs it is not a property of the system. The runbook is where the design meets the people carrying it out, and it can fail quietly by drifting from the system it describes.

A useful one states the decision as well as the steps: what evidence justifies failing over, what data loss is acceptable, and who may accept it (Failover is a decision, not an event). AWS asks for criteria the whole organisation understands. It lists what must already be in place, such as operator permissions in the standby Region, granted before going live rather than during the incident (The dependency you did not list). And it covers the way back, because the old primary may hold writes the new one lacks.

It is written for someone who may last have touched this service six months ago (A human in the loop adds minutes to recovery). Google's SRE book reports that a written playbook gives roughly three times better time to repair than improvising.

It stays correct only by being used. AWS recommends a single runbook for failover and failback, the same for tests and live events. At the FT, regular practice failovers caught changes the documentation had missed, which is the argument of Test the failover or you have not got one.