How Laravel queues fail without anyone noticing
A queue moves slow work, like emails, PDFs and calls to other systems, out of the web request. When it stops, the website still loads, so nobody sees it until a customer complains.
The workers died and nothing restarted them
How you'll know: Jobs pile up in the jobs table or in Redis, and nothing is processed until someone logs in and runs a command.
Fix: Run the workers under a process monitor such as Supervisor, as the Laravel docs describe, so a crashed worker comes straight back.
A deploy went out, but the workers run the old code
How you'll know: A fix works on the website but not in emails or background jobs, or jobs start throwing errors right after a release.
Fix: Workers are long-running processes that keep the code they started with. Add php artisan queue:restart to every deploy.
Failed jobs pile up where nobody looks
How you'll know: The failed_jobs table has hundreds of rows going back months.
Fix: List them with php artisan queue:failed, fix the cause, retry with queue:retry, and send an alert when new failures appear. Horizon gives a dashboard for this if you use Redis queues.
Long jobs run twice
How you'll know: Customers get two copies of an email or two charges, usually on the slowest jobs.
Fix: The worker timeout has to be shorter than the queue's retry_after setting, or a second worker picks the job up while the first is still running. Make jobs safe to run twice as well.
Scheduled tasks stopped after a server move
How you'll know: Nightly reports or reminders just stopped, around the time the app changed hosts.
Fix: The scheduler only runs if one cron entry calls php artisan schedule:run every minute. It's easy to leave off a new server.