node_exporter and Alertmanager interoperability on CentOS Stream: repeat_interval default uncertainty

On CentOS Stream systems deploying the documented Prometheus node_exporter and Alertmanager stack, operators must configure repeat_interval to balance re-notification against alert fatigue. node_exporter exports host metrics such as CPU, memory, and disk usage, while Alertmanager determines which alerts page on-call staff through grouping, inhibition, and silencing logic. The design goal is to set repeat_interval so that actionable alerts are re-sent without overwhelming recipients or masking subsequent failures.

A key uncertainty is that no documented default or platform-recommended value exists for repeat_interval in the CentOS or Prometheus ecosystems; the setting depends on organizational on-call practices rather than upstream defaults. Additionally, group_by scoping and inhibition rule interactions can affect whether repeated notifications are suppressed or duplicated when multiple alerts fire from the same host. Label mismatches between node_exporter metrics and Alertmanager matchers may also silently drop notifications, requiring careful testing.

What is the minimum repeat_interval that prevents alert suppression without triggering duplicate pages? How should group_by labels be scoped to avoid merging unrelated incidents while still supporting effective repeat_interval behavior? Can inhibition rules be configured to suppress repeated notifications of dependent alerts during a single host outage event?