I looked around as to what we can do to correct this but did not find anything helpful yet. I'm wondering why this has just suddenly popped up too. Most things I have found relate to kernel level stuff and because we have a vps, we don't seem to be able to manipulate any kernel parameters.It looks like there were a lot of stale TCP connections which caused a lock on numtcpsock in /proc/user_beancounters, and thus new TCP connections resulted in dead air. I've went ahead and raised that number to 300 from 160 so you should be ok for now.
Anyone out there have a cluestick to poke us with?