If etcd will hold cluster state, test it with a small production-shaped configuration and an explicit quorum plan. Write and read known keys, take one member offline, and verify the expected availability. Then rehearse snapshot creation and restore to a clean cluster.
Restrict access and monitor latency, storage growth, and member health. etcd is not a general-purpose database to adopt casually; the operational question is whether the team can preserve consensus and recover state during a failure.
Akousa