Every now and again I'll find that my NSCA server daemon deadlocks. The daemon itself is running but when you try to send it any data it doesn't do anything. Restarting fixes the issue. I haven't found a way to replicate the issue and happens very infrequently. Recently we had some networking issues which seemed to exasperate the issue causing this behavior to happen twice in one day.
We run the NSCA daemon in --single mode which I have found never times out a connection. What I thought was happening was that NSCA was using threads and it had a maximum number of threads it could spawn and during this connection issue we were having connections were being made that would never timeout consuming all the possible threads. As I found out looking at the code NSCA doesn't use threads at all when using --single but does asynchronous connection stuff instead.
I'm not adept enough at C, socket programming and especially asynchronous socket programming to know all the pitfalls but I'd imagine having a connection able to stay open forever and having that get polled is still a bad thing and can cause issues. Oh, and this is not an issue when running in any of the other modes. Seems to only happen in --single mode.