core: quarantine workers with sustained runtime storage stalls

infrastructure/core/kustomization.yaml

@@ -2,6 +2,7 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
+  - node-maintenance.yaml
   - ../modules/base
   - ../modules/profiles/atlas-ha
   - node-prefer-noschedule-serviceaccount.yaml

infrastructure/core/node-maintenance.yaml

@@ -0,0 +1,22 @@
+# infrastructure/core/node-maintenance.yaml
+# Block new application placement while the measured runtime I/O faults remain.
+# Existing workloads and storage DaemonSets are not drained by these resources.
+apiVersion: v1
+kind: Node
+metadata:
+  name: titan-14
+  annotations:
+    kustomize.toolkit.fluxcd.io/prune: disabled
+    atlas.bstein.dev/maintenance-reason: "Runtime USB flash saturation; replace or repair media and verify I/O before uncordoning."
+spec:
+  unschedulable: true
+---
+apiVersion: v1
+kind: Node
+metadata:
+  name: titan-18
+  annotations:
+    kustomize.toolkit.fluxcd.io/prune: disabled
+    atlas.bstein.dev/maintenance-reason: "Sustained runtime I/O stalls prevent Longhorn engine startup; reduce load and verify storage before uncordoning."
+spec:
+  unschedulable: true