Objectives before dashboards
Reliability work starts by agreeing what "working" means for each user journey, expressed as a service level objective. Without that, monitoring becomes a wall of graphs nobody can act on and alerting becomes noise.
Recovery objectives
Every workload gets an RPO — how much data loss is tolerable — and an RTO: how long recovery may take. These are business decisions, and they drive cost. Synchronous replication and a warm standby are not free, and not every system needs them.
Drills
A backup that has never been restored is an assumption. Recovery is exercised on a schedule, into an isolated environment, and the measured time is compared with the agreed RTO. When they disagree, one of them changes.