monitoring: delay PiManagerDown duplicate pages

apps/github-runner/README.md

@@ -24,6 +24,12 @@ original Longhorn ReadWriteOnce NuGet PVC. Every other repo-scoped runner uses
 two replicas with per-pod `emptyDir` caches. That is the safe backlog-drain
 strategy: no two pods share one RWO PVC.
 
+Ephemeral runner pods are expected to register, run one job, deregister, and
+exit so the Deployment starts a fresh pod for the next registration token. A
+small amount of exit-1/restart churn from token-expiry or no-work windows is
+accepted operational noise as long as jobs are not stuck queued and the
+repo-scoped runner-offline alerts stay quiet.
+
 Sprint 32 final long-tail wave adds 16 two-replica Deployments:
 `FlowerCore.Knowledge`, `FlowerCore.LlmBridge`, `FlowerCore.Media`,
 `FlowerCore.Presentations`, `FlowerCore.RemoteDesktop`, `FlowerCore.DNS`,

apps/monitoring/noc-monitoring.yaml

@@ -843,7 +843,9 @@ data:
         rules:
           - alert: PiManagerDown
             expr: up{job="pimanager-app"} == 0
-            for: 3m
+            # Sprint 67: delayed behind NodeDown's critical page so a powered-off
+            # Pi does not create the first duplicate page for the same host.
+            for: 8m
             labels:
               severity: warning
             annotations: