Accept delayed cAdvisor samples after transient scrape failures

flux/infrastructure/infra-monitor/helmrelease-prometheus.yaml

@@ -92,6 +92,9 @@ spec:
         podMonitorSelectorNilUsesHelmValues: false
         ruleSelectorNilUsesHelmValues: false
         scrapeInterval: 1m
+        tsdb:
+          # Retain delayed cAdvisor samples after a failed scrape inserted a stale marker.
+          outOfOrderTimeWindow: 2m
       ingress:
         enabled: false
         ingressClassName: nginx