fix hermes liveness probe

platform-engineer/deployment.yaml

@@ -128,12 +128,17 @@ spec:
             memory: "2Gi"
             cpu: "1000m"
         livenessProbe:
-          httpGet:
-            path: /health
-            port: 8642
-          initialDelaySeconds: 60
+          # Probe the dashboard port (9119, always enabled via HERMES_DASHBOARD=1
+          # and binds 0.0.0.0). The gateway API on 8642 is off by default
+          # (API_SERVER_ENABLED not set), so 9119 is the reliable liveness signal.
+          # s6 auto-restarts the gateway itself; this probe only catches a wedged
+          # container.
+          tcpSocket:
+            port: 9119
+          initialDelaySeconds: 90
           periodSeconds: 30
-          failureThreshold: 3
+          timeoutSeconds: 5
+          failureThreshold: 5
         securityContext:
           allowPrivilegeEscalation: false