Validate metric temporality end to end before trusting normal alerts

Metric ingestion can fail selectively by instrument type: runtime gauges may remain visible while counters and histograms are rejected because their delta/cumulative temporality is incompatible with the destination. Validate representative instruments through exporter, collector and query backend; a reachable collector or healthy gauge is not proof that error-rate and latency metrics arrived. Alert rules that treat missing data as OK can conceal this failure.

Applicability and limits: inspect explicit exporter errors and per-instrument query coverage before assigning this cause. Missing counters alone can also mean no activity or incorrect selectors. Correct temporality at the exporter or use a supported conversion processor, preserve per-instance identity, and verify observed recovery rather than assuming configuration acceptance proves ingestion.