🔧 Fix stale no_exit pid blocking restart loop 🔧

- Re-enter the monitor loop when /run/.no_exit.pid is left from a prior container life
- Only short-circuit when the recorded PID is still alive via kill -0
- Remove leftover pid file so the supervisor exec path runs on restart

.claude/scheduled_tasks.lock
rootfs/usr/local/etc/docker/functions/entrypoint.sh

.claude/scheduled_tasks.lock

@@ -0,0 +1 @@
+{"sessionId":"e3874c4a-b4c6-45a3-a533-a31e6bb59f0e","pid":3227263,"procStart":"144242774","acquiredAt":1778024153040}

rootfs/usr/local/etc/docker/functions/entrypoint.sh

@@ -280,7 +280,16 @@ __no_exit() {
   local failed_services=""
   local failure_count=0
 
-  [ -f "/run/.no_exit.pid" ] && return 0
+  # only return early if the recorded PID is still alive; a leftover
+  # pid file from a prior container life (docker restart) would otherwise
+  # cause us to exit instead of entering the monitor loop.
+  if [ -f "/run/.no_exit.pid" ]; then
+    no_exit_pid="$(cat /run/.no_exit.pid 2>/dev/null)"
+    if [ -n "$no_exit_pid" ] && kill -0 "$no_exit_pid" 2>/dev/null; then
+      return 0
+    fi
+    rm -f /run/.no_exit.pid 2>/dev/null || true
+  fi
 
   exec bash -c "
     trap 'echo \"Container shutdown requested\"; rm -f /run/.no_exit.pid /run/*.pid; exit 0' TERM INT