I saw how this happens, I think that you are looking in the right general direction.
When this happens, I assume that you can accurately determine the exact equipment? after failure, I recommend running memtest (http://www.memtest.org/).
I have seen such things with power supplies, bad disk controllers, etc. You can go crazy trying to track them.
It seems you are going the right way about it - if you can find a way to make the problem happen more quickly when it happens, run MEMTEST, run CHKDSK / R (check EventLog for controller errors during this)
any chance that you can hook up a kernel debugger?
Is there any chance of% SystemRoot% \ memory.dmp?
source share