The jobs table keeps growing, nothing fails, no error is logged - and the queue worker looks perfectly alive. This is the most common queue incident in Laravel, and it is almost never mysterious once you know the four places a worker silently stops: stale code, the memory ceiling, a misconfigured retry_after, and a missing process manager.
1. The worker is running old code
queue:work is a daemon: it boots the framework once and keeps the application in memory. Deploy new code and the worker keeps executing the old code. Worse, if the deploy changed a job class or database schema, the old worker can freeze or mishandle new payloads without a clear error.
Fix: restart workers on every deploy.
php artisan queue:restart
This writes a timestamp to the cache; each worker exits gracefully after its current job, and Supervisor (below) starts a fresh one on the new code. Add it to your deploy script permanently - this is the single most forgotten deploy step in Laravel.
2. The memory limit killed it quietly
Workers are long-lived, and PHP leaks a little memory per job. Start workers with a ceiling:
php artisan queue:work --memory=128
When a worker crosses 128 MB it finishes its current job and exits - by design. If nothing restarts it, the queue simply stops. The exit is intentional and clean, so no error appears anywhere. Pair the memory limit with Supervisor's autorestart=true and this behavior becomes self-healing instead of fatal. The flags --max-jobs=1000 and --max-time=3600 recycle workers on the same principle.
3. retry_after shorter than the job timeout
In config/queue.php, retry_after decides when a reserved job is considered lost and put back on the queue. If it is shorter than the worker's --timeout, a long job gets released back while still running - then processed twice, or bounced between workers forever, looking like a stalled queue:
// config/queue.php - WRONG
'retry_after' => 60, // job released after 60s
// worker started with --timeout=120 ... job runs 120s
Rule: retry_after must exceed timeout by a comfortable margin (e.g. timeout 60, retry_after 90).
4. Nothing is keeping the worker alive
A worker started by hand in an SSH session dies with the session. Even a cron or nohup worker dies on OOM, timeout, or its memory limit. Supervisor is the standard babysitter:
[program:laravel-worker]
process_name=%(program_name)s_%(process_num)02d
command=php /var/www/app/artisan queue:work --sleep=3 --tries=3 --memory=128
autostart=true
autorestart=true
user=www-data
numprocs=2
redirect_stderr=true
stdout_logfile=/var/log/laravel-worker.log
Check its view of the world first: sudo supervisorctl status. A worker in FATAL or BACKOFF state is your stalled queue.
When it still looks stuck
Confirm the basics: jobs landing in failed_jobs (queue:failed), the queue connection in .env matching what the worker reads (config:clear after env changes!), and the worker listening on the same queue name the jobs are pushed to (queue:work --queue=high,default vs dispatch()->onQueue('emails')). If you run multiple servers or need a dashboard, Laravel Horizon wraps all of this with auto-balancing and metrics.
Summary
A silent queue is a lifecycle problem, not a code bug. Restart workers on every deploy, cap memory with --memory, keep retry_after above timeout, and let Supervisor own restarts. Do those four and the queue stops being the thing you check first every morning.