Is this repo still active, and is this still a problem?
What have you discussed to approach solving?
Some ideas I have:
Heartbeat & TTL based shutdown.
So long as you heartbeat multiple intervals between ttl, you could have the servers self-shutdown, on what I might call a tick (iteration in a loop); so that the language server instance knows if it should be live.
Max runtime / requests
This borrows an idea from the PHP-FPM server gateway project; likely others. You could optionally remove the server after {N} requests, or during any exceptional events. The cost here is time to spin-up a new language server, so it's opinionated as a choice, but simpler than a heartbeat TTL; and could even be a step towards that goal, as being able to shut-down and boot at subsequent requests underpins stopping servers on disconnect / inactivity.
Is this repo still active, and is this still a problem?
What have you discussed to approach solving?
Some ideas I have:
Heartbeat & TTL based shutdown. So long as you heartbeat multiple intervals between ttl, you could have the servers self-shutdown, on what I might call a tick (iteration in a loop); so that the language server instance knows if it should be live.
Max runtime / requests This borrows an idea from the PHP-FPM server gateway project; likely others. You could optionally remove the server after {N} requests, or during any exceptional events. The cost here is time to spin-up a new language server, so it's opinionated as a choice, but simpler than a heartbeat TTL; and could even be a step towards that goal, as being able to shut-down and boot at subsequent requests underpins stopping servers on disconnect / inactivity.