Pointing monit at a process that keeps dying makes the symptom disappear, and that is precisely the risk: the service comes back within thirty seconds, the graph looks fine, and nobody investigates why it needed to.
check process queue-worker with pidfile /var/run/worker.pid
start program = "/usr/sbin/service worker start"
stop program = "/usr/sbin/service worker stop"
if 5 restarts within 5 cycles then alert
if memory > 200 MB for 3 cycles then restartThe line that earns its keep is if 5 restarts within 5 cycles then alert — it distinguishes a process that fell over once from one that is crash-looping, and only the second is worth waking someone for. Restarting on a memory threshold is a legitimate mitigation for a known leak in a dependency; it is not a fix, and the alert is what stops it from being mistaken for one.