Docker-compose setup

I have successfully setup a couple of libretime instances using the docker-compose procedure. Works great.
I am struggling with some reliability issues with the libretime-liquidsoap instance. It seems very similar to this “Can not connect to the streaming server. Problem with Liquidsoap · Issue #1606 · libretime/libretime · GitHub” but I don’t understand the references and jargon in that dialog.
( using libretime version 3.0.2 )

From the logs :

liquidsoap_1 | 2023/05/09 16:22:33 [LibreTime!:2] Error while sending data: could not write data to host: Broken pipe in write()!
liquidsoap_1 | 2023/05/09 16:22:33 [lang:3] timeout --signal=KILL 45 libretime-playout-notify stream ‘1’ ‘1683569903.49’ --error=‘could not write data to host: Broken pipe in write()’ &
liquidsoap_1 | 2023/05/09 16:22:33 [LibreTime!:3] Closing connection…
liquidsoap_1 | 2023/05/09 16:22:33 [LibreTime!:3] Will try to reconnect in 5.00 seconds.
liquidsoap_1 | 2023-05-09 16:22:34.708 | INFO | libretime_playout.notify.main:stream:92 - Sending output stream ‘1’ status ‘could not write data to host: Broken pipe in write()’
liquidsoap_1 | 2023/05/09 16:22:39 [LibreTime!:3] Connecting mount nhsw0ddddd for source@xxxxx…
liquidsoap_1 | 2023/05/09 16:22:39 [LibreTime!:3] Connection setup was successful.

The end result seems to be that a lot of zombie liquidsoap processes are being created and the only option is restarting from time to time.

1000 5383 5357 8 May08 ? 01:54:20 | _ libretime-liquidsoap --verbose /app/radio.liq
1000 9131 5383 0 May08 ? 00:00:00 | _ [timeout]
1000 9145 5383 0 May08 ? 00:00:00 | _ [timeout]
1000 18508 5383 0 May08 ? 00:00:00 | _ [timeout]
1000 19129 5383 0 May08 ? 00:00:00 | _ [timeout]
1000 5695 5383 0 May08 ? 00:00:00 | _ [timeout]
1000 6125 5383 0 May08 ? 00:00:00 | _ [timeout]
( looots more )

( you may tell I am new to libretime but somewhat familiar with liquidsoap and icecasts and whatnot … )

Any advice appreciated.

Maybe liquidsoap cannot send data to icecast. Could paste more logs, from playout and from icecast ?

It was indeed on the icecast side - it was force closing idle ( no listener ) connections. I just assumed that part was good.
Thank you for the pointer.
It will still be good not to leak those defunct processes even if that situation occurs for some network issues .
I think this line may be the culprit:
command = “timeout --signal=KILL 45 libretime-playout-notify #{args} &”
I guess a liquidsoap issue.