diff --git a/Watchdogs/stability_watchdog.sh b/Watchdogs/stability_watchdog.sh
index 0c7f4ad..749b648 100755
--- a/Watchdogs/stability_watchdog.sh
+++ b/Watchdogs/stability_watchdog.sh
@@ -535,7 +535,28 @@ do_reboot() {
     if is_docker_enabled && command -v docker >/dev/null 2>&1; then
         mapfile -t _SYS_REBOOT_STOPPED < <(timeout "$DOCKER_TIMEOUT" docker ps --format '{{.Names}}' 2>/dev/null)
         trap _trap_sys_reboot_restart EXIT
-        timeout 60 docker ps -q 2>/dev/null | xargs -r docker stop >/dev/null 2>&1
+
+        # One container at a time, bounded individually, named when it fails.
+        #
+        # This was `timeout 60 docker ps -q | xargs -r docker stop`, where the timeout bounded
+        # only the left of the pipe — the stops themselves were unbounded, so a wedged daemon
+        # hung the reboot path indefinitely, on the half that actually stops the containers.
+        #
+        # Bounding the batch instead would mean picking a number against the container count,
+        # and being wrong either kills containers mid-write or never returns. Per container that
+        # question disappears: DOCKER_TIMEOUT is three times docker's own 10s stop grace, and
+        # holds whether this host runs five containers or fifty.
+        #
+        # It also reuses the list already read above rather than asking docker a second time,
+        # and matches stop_local_containers() in common.sh — loop, report by name, never xargs.
+        for _c in "${_SYS_REBOOT_STOPPED[@]}"; do
+            [[ -z "$_c" ]] && continue
+            if timeout "$DOCKER_TIMEOUT" docker stop "$_c" >/dev/null 2>&1; then
+                log "  $_c stopped"
+            else
+                warn "  $_c did not stop within ${DOCKER_TIMEOUT}s — continuing to reboot"
+            fi
+        done
     fi
 
     warn "Stopping User Scripts..."