Description of problem:

When using multiple workers, we might loose metrics. This seems to be caused by a race-condition causing the different workers to federate the same metrics from Prometheus.

Version-Release number of selected component (if applicable):

ACM 2.12-ACM2.15

How reproducible:

• Most of the time

Steps to Reproduce:

1. Setup a ACM Hub+spoke
2. Scale up to use more than 1 worker in the observabilityAddonSpec in the MCO CR
3. You may need to restart metric-collector a few times to hit bad state
4. You can view the count( {cluster="your-cluster-name"}

   ) to confirm the issue. It should be very stable and not change when the metric collector is restarted

Actual results:

• Metrics might be missing
• The easiest way to confirm this issue if when the multiple workers are sending the exact same number of timeseries, like below (it also seem to always happen on startup):

level=debug caller=logger.go:45 ts=2025-10-03T13:45:13.278299233Z shard=3 component=forwarder component=metricsclient timeseriesnumber=13730
level=debug caller=logger.go:45 ts=2025-10-03T13:45:13.318888053Z shard=0 component=forwarder component=metricsclient timeseriesnumber=13730
level=debug caller=logger.go:45 ts=2025-10-03T13:45:13.320064215Z shard=1 component=forwarder component=metricsclient timeseriesnumber=13730
level=debug caller=logger.go:45 ts=2025-10-03T13:45:13.367873111Z shard=2 component=forwarder component=metricsclient timeseriesnumber=15300

Expected results:

• No metrics are lost. The workers produce a different number of timeseries.

Additional info: