1. Proposed title of this feature request:
Allow metric-specific labeldrop configuration in metric_relabel_configs of ServiceMonitors/PodMonitors

2. What is the nature and description of the request?
Currently, when using metric_relabel_configs in ServiceMonitors/PodMonitors, applying a labeldrop (e.g., to remove pod or instance labels) affects all collected metrics indiscriminately. The request is to enable targeting specific metrics so that labels can be dropped only from selected metrics, not globally across all.

3. Why does the customer need this? (List the business requirements here)
The customer has long-running alerts tied to specific metrics. Pod names frequently change (due to restarts, upgrades, etc.), which changes the pod label value. Since the pod label is part of the alert fingerprint in AlertManager, these changes result in a new alert fingerprint, causing:

• AlertManager to treat it as a new alert.

• Triggering of redundant alerts and webhook notifications.

• Difficulty in tracking the same logical alert across restarts.

• Manual overhead to reconcile related alerts.

To prevent this, the customer wants to drop the pod label only from alert-critical metrics to preserve alert fingerprint consistency, while still keeping the label for other metrics where pod identity is valuable for observability.

4. List any affected packages or components.

• OpenShift Monitoring Stack

• Prometheus Operator

• ServiceMonitor / PodMonitor

• metric_relabel_configs