diff --git a/deploy/gate_log.jsonl b/deploy/gate_log.jsonl
index 285c1c7f..e32516b5 100644
--- a/deploy/gate_log.jsonl
+++ b/deploy/gate_log.jsonl
@@ -426,3 +426,13 @@
 {"ts":"2026-06-07T20:17:03+08:00","date":"2026-06-07","gate":"shell_eol","result":"PASS","detail":"all tracked sh LF"}
 {"ts":"2026-06-07T20:17:03+08:00","date":"2026-06-07","gate":"text_eol","result":"PASS","detail":"git index LF"}
 {"ts":"2026-06-07T20:17:03+08:00","date":"2026-06-07","gate":"summary","result":"PASS","detail":"7/7"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"changelog","result":"PASS","detail":"2026-06-07-health-monitor-flock-fail-count-fix.md 21lines"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"audit","result":"PASS","detail":"2026-06-07-ops-patrol-layer3-telegram.md"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"test","result":"PASS","detail":"all passed"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"lint","result":"PASS","detail":"0 violations"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"import","result":"PASS","detail":"ok"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"security","result":"PASS","detail":"0 HIGH"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"review","result":"PASS","detail":"audit covers changes"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"shell_eol","result":"PASS","detail":"all tracked sh LF"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"text_eol","result":"PASS","detail":"git index LF"}
+{"ts":"2026-06-07T20:53:00+08:00","date":"2026-06-07","gate":"summary","result":"PASS","detail":"7/7"}
diff --git a/deploy/health_monitor.sh b/deploy/health_monitor.sh
index 7cdb1bff..871236a5 100644
--- a/deploy/health_monitor.sh
+++ b/deploy/health_monitor.sh
@@ -106,6 +106,7 @@ bump_fail_count() {
             fail=$((fail + 1))
             echo "$fail" > "$FAIL_FILE"
         ) 9>"${FAIL_FILE}.lock"
+        fail="$(cat "$FAIL_FILE" 2>/dev/null || echo 0)"
     else
         fail="$(cat "$FAIL_FILE" 2>/dev/null || echo 0)"
         fail=$((fail + 1))
diff --git a/docs/audit/2026-06-07-ops-patrol-layer3-telegram.md b/docs/audit/2026-06-07-ops-patrol-layer3-telegram.md
index da513daf..ae135cb1 100644
--- a/docs/audit/2026-06-07-ops-patrol-layer3-telegram.md
+++ b/docs/audit/2026-06-07-ops-patrol-layer3-telegram.md
@@ -7,7 +7,7 @@
 | `server/infrastructure/env_file.py` | `.env` upsert/read（UTF-8 LF） |
 | `server/application/services/ops_patrol_sync.py` | Telegram → `/app/.env`、persist 卷、host `.env.prod` |
 | `server/api/settings.py` | `set_setting` 钩子；`GET/POST ops-patrol/*` |
-| `deploy/health_monitor.sh` | `load_telegram_from_db` MySQL 回退 |
+| `deploy/health_monitor.sh` | `load_telegram_from_db` + `bump_fail_count` flock 修复 |
 | `deploy/deploy-production.sh` | 部署成功自动 `install_ops_cron.sh` |
 | `deploy/nexus-1panel.sh` | `sync_env_prod_to_volume` 含 Telegram 键 |
 | `frontend/src/pages/SettingsPage.vue` | Layer 3 巡检状态与同步 UI |
diff --git a/docs/changelog/2026-06-07-health-monitor-flock-fail-count-fix.md b/docs/changelog/2026-06-07-health-monitor-flock-fail-count-fix.md
new file mode 100644
index 00000000..123a0ecb
--- /dev/null
+++ b/docs/changelog/2026-06-07-health-monitor-flock-fail-count-fix.md
@@ -0,0 +1,21 @@
+# Changelog — health_monitor flock 失败计数修复
+
+**日期**: 2026-06-07
+
+## 摘要
+
+修复 `bump_fail_count` 在使用 `flock` 时子 shell 未回传计数、导致 `FAIL` 恒为 0、Layer 3 永不触发重启/Telegram 的 bug。
+
+## 动机
+
+生产模拟巡检时发现连续 3 次 /health 失败后 `last_restart` 仍为空；`bash -x` 显示 `FAIL=0` 而 `/tmp/nexus_health_fail_count` 已为 3。
+
+## 涉及文件
+
+- `deploy/health_monitor.sh`
+
+## 验证
+
+```bash
+# 阻断 127.0.0.1:8600 后连续执行 3 次 health_monitor.sh，第 3 次应触发 Telegram + docker restart
+```