Occasionally pm2 managed script will error and pm2 will attempt to automatically restart, but forked child processes are not getting killed. How can I make sure they are killed before pm2 attempts to restart the main script?

I use pm2 to run my custom-written media server on my server.
It runs round the clock all day, and it's difficult to test locally for such long periods. So every now and then it will error with some bad code I've written, which I will promptly fix.

I want it to restart upon erroring, which it does...
But my script also calls child_process.fork on a few other scripts to run under my main app, and these processes need to communicate between one another with IPC messaging.

When I encounter an error in the main process, the child processes get disconnected from the main process and remain running. So pm2 tries to restart the main script (and one of these forked scripts requires a specific port not to be occupied), then it runs into trouble again.

I have to manually end the floating processes and then run pm2 restart <script>

So what's the solution? (besides write better code)

I read that process's uncaughtException event should only be used as a last resort... which I believe this is - so if I set up a listener in the main script with a function to terminate the child processes with force, will pm2 wait for them to close before starting the main process again?

Start by reading PM2's restart and child-process handling behavior alongside Node.js child_process.fork and IPC documentation. Reproduce the failure with a forked child that holds a port, then verify that all child processes are terminated and the port is released before PM2 restarts the main script.