That's strange. I have also been running the OC'd Titan kernel for quite a while without any problems, and yesterday I had the same thing grog just described. I was probably using GRead too (I always do :-). I thought it was maybe a memory issue and a watchdog kicking in.

Grog, are you also using some of debernardis' "optimizations" scripts ?

One of those says that when a memalloc fails, kill the process that asked for it, not the one that's using the most memory... so if a user process is super greedy, but the failed memalloc happens to come from a system process, maybe that could trigger a reboot ?