Recovering the wrong thing first wastes time.
The service graph determines the sequence: restore what everything else depends on, then bring the visible application back.
Network + accessRouting, IAM, secrets, certificates.
Core dataDatabase, queues, object storage.
Platform servicesIdentity, observability, messaging.
User entry pointsAPIs, frontends, ingress, DNS.
Communication path
Status must survive prod
A status page or incident room hosted in the same failed environment is not a communication plan.
Observability path
Monitoring must survive prod
If dashboards disappear with the outage, recovery quality drops fast.