diff --git a/charts/lobu/templates/prometheusrule.yaml b/charts/lobu/templates/prometheusrule.yaml
index e6696dbd2..239fdee62 100644
--- a/charts/lobu/templates/prometheusrule.yaml
+++ b/charts/lobu/templates/prometheusrule.yaml
@@ -23,7 +23,7 @@ spec:
         # replicas down / never scraped).
         - alert: WatcherAutomationSilent
           expr: |
-            (sum(increase(lobu_scheduled_job_runs_total{job="watcher-automation",outcome="success"}[15m])) or on() vector(0)) == 0
+            (sum(increase(lobu_scheduled_job_runs_total{task="watcher-automation",outcome="success"}[15m])) or on() vector(0)) == 0
           for: 15m
           labels:
             severity: critical
@@ -46,12 +46,12 @@ spec:
         # A critical scheduled job erroring repeatedly (e.g. the stale-run reaper).
         - alert: LobuScheduledJobFailing
           expr: |
-            sum(increase(lobu_scheduled_job_runs_total{job=~"watcher-automation|check-stalled-executions",outcome="error"}[15m])) by (job) > 0
+            sum(increase(lobu_scheduled_job_runs_total{task=~"watcher-automation|check-stalled-executions",outcome="error"}[15m])) by (task) > 0
           for: 15m
           labels:
             severity: warning
             service: lobu
           annotations:
-            summary: "Lobu scheduled job {{`{{ $labels.job }}`}} is failing"
-            description: "The {{`{{ $labels.job }}`}} cron task threw on every run over 15m."
+            summary: "Lobu scheduled job {{`{{ $labels.task }}`}} is failing"
+            description: "The {{`{{ $labels.task }}`}} cron task threw on every run over 15m."
 {{- end }}
diff --git a/charts/lobu/templates/servicemonitor.yaml b/charts/lobu/templates/servicemonitor.yaml
index 8c093bb76..cd5e72add 100644
--- a/charts/lobu/templates/servicemonitor.yaml
+++ b/charts/lobu/templates/servicemonitor.yaml
@@ -23,8 +23,11 @@ spec:
     matchLabels:
       {{- include "lobu.appSelectorLabels" . | nindent 6 }}
   endpoints:
+      # The Prometheus exporter lives on the gateway, which the unified server
+      # mounts under /lobu — so metrics are at /lobu/metrics, not /metrics
+      # (root /metrics 302-redirects to the SPA → scrape fails on Content-Type).
     - port: http
-      path: /metrics
+      path: {{ .Values.metrics.serviceMonitor.path }}
       interval: {{ .Values.metrics.serviceMonitor.interval }}
       scrapeTimeout: {{ .Values.metrics.serviceMonitor.scrapeTimeout }}
 {{- end }}
diff --git a/charts/lobu/values.yaml b/charts/lobu/values.yaml
index a4c9df6de..51d10d1ef 100644
--- a/charts/lobu/values.yaml
+++ b/charts/lobu/values.yaml
@@ -229,6 +229,9 @@ metrics:
   serviceMonitor:
     enabled: false
     # namespace: ""        # defaults to the release namespace
+    # Metrics are exposed by the gateway, mounted under /lobu by the unified
+    # server — so the scrape path is /lobu/metrics (root /metrics redirects).
+    path: /lobu/metrics
     interval: 30s
     scrapeTimeout: 10s
     additionalLabels: {}   # e.g. { release: kube-prometheus-stack }
diff --git a/packages/server/src/gateway/metrics/prometheus.ts b/packages/server/src/gateway/metrics/prometheus.ts
index 92a7598ee..d6bd3c562 100644
--- a/packages/server/src/gateway/metrics/prometheus.ts
+++ b/packages/server/src/gateway/metrics/prometheus.ts
@@ -72,7 +72,7 @@ function initializeMetrics() {
   // handled by rate()/increase().
   registerMetric(
     "lobu_scheduled_job_runs_total",
-    "Scheduled (cron) task ticks by job name and outcome (success|error)",
+    "Scheduled (cron) task ticks by task name and outcome (success|error)",
     "counter"
   );
   registerMetric(
diff --git a/packages/server/src/scheduled/task-scheduler.ts b/packages/server/src/scheduled/task-scheduler.ts
index 15063fd09..fbed69f6d 100644
--- a/packages/server/src/scheduled/task-scheduler.ts
+++ b/packages/server/src/scheduled/task-scheduler.ts
@@ -256,13 +256,16 @@ export class TaskScheduler {
         payload: data.payload,
         taskRunId: Number(job.id),
       });
+      // Label is `task`, NOT `job`: Prometheus reserves `job` for the scrape
+      // target and overwrites any same-named metric label, which silently
+      // collapses per-task series (and broke the watcher-automation alert).
       incrementCounter('lobu_scheduled_job_runs_total', {
-        job: data.name,
+        task: data.name,
         outcome: 'success',
       });
     } catch (err) {
       incrementCounter('lobu_scheduled_job_runs_total', {
-        job: data.name,
+        task: data.name,
         outcome: 'error',
       });
       throw err;