As far as I can tell, the only way to rotate logs is to issue a restart of the service, which causes all clients to disconnect, it also initiates a full recheck of the npm and dependencies, which slows things down (e.g. /etc/init.d/etherpad-lite restart).

It would be ideal if we could do a 'kill -HUP' to a process to cause it to close the log file handle and open a new one, as this is the least disruptive mechanism for rotating logs, and is pretty common for services to use.

I tried to do a kill -HUP on the server.js process, but that just killed things outright.