Evaluate Need for Cloud SQL Read Replicas Across Projects

Some projects currently have Cloud SQL read replicas configured, but it’s unclear whether they are actually needed or being used.

Observations so far:

Many instances are Enterprise (non-Enterprise Plus), so features like cross-region DR replicas or read pools don’t apply.
Read replicas do not automatically route reads. Jobs and applications require separate connection strings to use them.
Most jobs perform writes, so they cannot safely use a replica.

No documented services or workloads are currently configured to read from these replicas.
The apparent use case may be for manual/ad-hoc queries, but this isn’t confirmed. Read replicas effectively double storage costs.

Which read replicas are actually being used, and by whom?
Are they intended only for manual queries?
Do any workloads require read scaling or DR for these replicas?

Evaluate each read replica across projects for actual usage and value.

Decide whether to:

Keep (and document usage/connection strategy)

Remove (to reduce cost and complexity)