diff --git a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicator.kt b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicator.kt
index 2715609..6498d70 100644
--- a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicator.kt
+++ b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicator.kt
@@ -45,6 +45,7 @@ internal fun jobRunnerHealth(snapshot: JobActivity.Snapshot, now: Instant): Heal
                 "started" to snapshot.started,
                 "lastTickAgeSeconds" to (ageSeconds ?: -1L),
                 "lastTickDurationMs" to snapshot.lastTickMillis,
+                "lastFailureAgeSeconds" to (snapshot.lastFailureAt?.let { Duration.between(it, now).seconds } ?: -1L),
                 "ticks" to snapshot.ticks,
                 "failures" to snapshot.failures,
                 "lastSweepSelected" to snapshot.lastSweepSelected,
diff --git a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/JobActivity.kt b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/JobActivity.kt
index bfabbd6..77eff44 100644
--- a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/JobActivity.kt
+++ b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/JobActivity.kt
@@ -22,6 +22,9 @@ class JobActivity {
     @Volatile
     private var lastTickMillis = 0L
 
+    @Volatile
+    private var lastFailureAt: Instant? = null
+
     /** 上一轮回填扫描选中的待办条数（扫描积压的即时值）。 */
     @Volatile
     var lastSweepSelected = 0
@@ -46,18 +49,20 @@ class JobActivity {
         ticks.incrementAndGet()
     }
 
-    /** 一次作业 tick 抛错：只累计失败，**不更新心跳时刻**——停摆由心跳年龄暴露。 */
-    fun tickFailed() {
+    /** 一次作业 tick 抛错：记下最近失败时刻，**不更新心跳时刻**——停摆由心跳年龄暴露。 */
+    fun tickFailed(now: Instant) {
+        lastFailureAt = now
         failures.incrementAndGet()
     }
 
     fun snapshot(): Snapshot =
-        Snapshot(started, lastTickAt, lastTickMillis, lastSweepSelected, ticks.get(), failures.get())
+        Snapshot(started, lastTickAt, lastTickMillis, lastFailureAt, lastSweepSelected, ticks.get(), failures.get())
 
     data class Snapshot(
         val started: Boolean,
         val lastTickAt: Instant?,
         val lastTickMillis: Long,
+        val lastFailureAt: Instant?,
         val lastSweepSelected: Int,
         val ticks: Long,
         val failures: Long,
diff --git a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/PipelineMetrics.kt b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/PipelineMetrics.kt
index 2259799..ff1a85d 100644
--- a/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/PipelineMetrics.kt
+++ b/src/main/kotlin/com/gzzn/omms/msgexchange/infra/metrics/PipelineMetrics.kt
@@ -24,6 +24,7 @@ import java.time.Duration
  * - `msgx.pipeline.watermark.lag`：水位落后信箱最新 ID 的距离
  * - `msgx.pipeline.hole.aged_out.total`：永久空洞放行次数
  * - `msgx.pipeline.job.heartbeat_age_seconds`：距上一次作业 tick 完成的秒数（未跑过为 -1）
+ * - `msgx.pipeline.job.last_failure_age_seconds`：距最近一次作业 tick 失败的秒数（从未失败为 -1）
  * - `msgx.pipeline.job.ticks.total` / `msgx.pipeline.job.failures.total`：作业 tick 完成/抛错次数
  * - `msgx.pipeline.job.last_sweep_selected`：上一轮回填扫描选中的待办条数（扫描积压）
  *
@@ -84,6 +85,10 @@ class PipelineMetrics(
             a.snapshot().lastTickAt?.let { Duration.between(it, clock.instant()).seconds.toDouble() } ?: -1.0
         }.strongReference(true).register(registry)
 
+        Gauge.builder("msgx.pipeline.job.last_failure_age_seconds", activity) { a ->
+            a.snapshot().lastFailureAt?.let { Duration.between(it, clock.instant()).seconds.toDouble() } ?: -1.0
+        }.strongReference(true).register(registry)
+
         Gauge.builder("msgx.pipeline.job.ticks.total", activity) { it.snapshot().ticks.toDouble() }
             .strongReference(true)
             .register(registry)
diff --git a/src/main/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunner.kt b/src/main/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunner.kt
index 5dd75e0..762cd62 100644
--- a/src/main/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunner.kt
+++ b/src/main/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunner.kt
@@ -79,7 +79,7 @@ class JobRunner(
         } catch (e: InterruptedException) {
             throw e
         } catch (e: Exception) {
-            activity.tickFailed()
+            activity.tickFailed(startedAt)
             log.warn("job tick failed: {}", e.message ?: e.javaClass.simpleName)
         }
     }
diff --git a/src/test/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicatorTest.kt b/src/test/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicatorTest.kt
index 9290b51..bca15f6 100644
--- a/src/test/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicatorTest.kt
+++ b/src/test/kotlin/com/gzzn/omms/msgexchange/infra/health/JobRunnerHealthIndicatorTest.kt
@@ -59,12 +59,16 @@ class JobRunnerHealthIndicatorTest {
     fun `a failed tick counts up without refreshing the heartbeat`() {
         val activity = JobActivity().apply { started() }
 
-        activity.tickFailed()
-        activity.tickFailed()
+        activity.tickFailed(now.minusSeconds(5))
+        activity.tickFailed(now)
 
         val snapshot = activity.snapshot()
         assertEquals(2L, snapshot.failures)
         assertEquals(0L, snapshot.ticks)
         assertEquals(null, snapshot.lastTickAt)
+
+        val details = details(jobRunnerHealth(snapshot, now))
+        assertEquals(0L, details["lastFailureAgeSeconds"])
+        assertEquals(2L, details["failures"])
     }
 }
diff --git a/src/test/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunnerTest.kt b/src/test/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunnerTest.kt
index 7c94785..dda6a71 100644
--- a/src/test/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunnerTest.kt
+++ b/src/test/kotlin/com/gzzn/omms/msgexchange/jobs/JobRunnerTest.kt
@@ -69,6 +69,7 @@ class JobRunnerTest {
 
         val snapshot = activity.snapshot()
         assertNull(snapshot.lastTickAt)
+        assertEquals(MutableClock.BASE, snapshot.lastFailureAt)
         assertEquals(0L, snapshot.ticks)
         assertEquals(1L, snapshot.failures)
     }