pg_static checking targets in a series rather than parallel

What did you do?
We run Postgres_Exporter in deployments in K8 targeting AWS/RDS endpoints. Each deployment targets several Postgres instances by their major version. We use the metric pg_static to test for availability. We have Alert Rules set up in Grafana to check UP/DOWN.

What did you expect to see?
If one Postgres instance is stopped then all others configured in the same deployment should still show available.

What did you see instead? Under which circumstances?
All Postgres instances/targets in the deployment showed "no data" and as a result alerts were triggered for database down. This remained until a restart of the deployment and/or a removal of the offending target.

Environment
AWS/RDS, Postgres v11, v12 and v13.

• System information:

AWS K8 pod

• postgres_exporter version:
  v0.10.1

• postgres_exporter flags:

None

• PostgreSQL version:

v11.9, 11.13, 11.16, 12.8, 13.4, 13.6, 13.7

• Logs:

Will follow up with these when available

No source file or test is identified in the issue. Start by reproducing the deployment with several PostgreSQL targets and one unavailable instance, then trace the pg_static collection path. Done means healthy targets still report availability when one target is stopped, and Grafana alerts distinguish each target correctly.