[SR-Users] Detecting problems of kamailio under very high load

We have multiple kamailio servers with 4 cpu cores and 16G RAM. We use kamailio+rtpproxy as a outbound sip proxy. Usually there are many thousands of concurrent sip sessions of occurring there. Periodically sometimes it just stops serving request and spits out 5** replies. At that point we usually reload the kamailio daemon and things return back to normal. As we have a lot of servers, manually doing this is a pain in the neck. So We installed a homer/sipcapture on a separate server . And from there, we periodically scan for 500 msgs for a given time interval. If there are any, we then proceed to reload kamailio. I know its hell of an inefficient system to monitor. So I'm wondering, if there can be any more rational way to detect problems pre-emptively. any fifo/mi command which we could run to find out if the kamailio instance is hung up? Cheers Arif