Hi jsteffen,
I tried upgrading from ESXi v4.0 to update 1 and lastly to update 2, where the problems still persisted.
Problems for me ranged from spontaneous reboot to hangs with nothing on the screen and nothing on a keyboard would wake it up (Num-Lock didn't even toggle, so unlikely to be software related).
Lastly, I got some approved down time and shifted the CPU from one host (that failed) to another (that was working fine) and the failures happened on the same system irrespective of which CPU was in there, indicating that it wasn't CPU related. I then shifted VMs off to the system that was working and tried running memtest v4.0 on the system to check the RAM. 3-and-a-bit iterations later it was reporting all was fine. There's not a lot left in the system, so I sent the motherboard back under warranty.
To give another board a boost, I put the RAM and CPU from the failing system into it, while the board was off under RMA. Within 3 days the beefier system was now failing. Repeatedly. Now, knowing the CPU was fine, the introduction of the RAM was also introducing faults into a system that was otherwise fine. 3 iterations of memtest didn't show any errors, and it took 3 days for one to be found, where it failed 3 times in the same day. Talk about intermittent.
Taking the additional RAM out, and the system hasn't shown problems in 30+ hours now. Still waiting.
I don't know if this will help you, but it appears that the PSOD (purple screen of death) is related to a hardware problem. Not bad for ECC Reg server class RAM. No reported ECC faults that I can see... Maybe I'm just not looking in the right place. Maybe if it is reported, there needs to be a "configure reporting" tab in vSphere...
