Ok, given that your saying first runs off a cold start tend to be fine, and then, right when it's sucked you in to thinking everything is good again, it kicks you in the nuts: this may indicate something like heat soak.
The main thing that changes the longer a machine is left running is heat, my guess is something no longer tolerates heat like it used to and becomes unstable. I now everything appears to be within specs temp wise, but components can change their tolerance for heat/voltage/amps once they have degraded, particulalry if there was a non critical fault there originally.
Given that your saying the 260 GTX can replicate the problem as well, you can eliminate the 295 GPU's as being the problem.
You've also swapped out the PSU and replicated the problem, which means its unlikely both psu's are having heat problems giving the exact same symptoms.
You've changed OS that many farkin times it not OS related.
You've swapped CPU as well and also replicated the borkedness (new word I made up
), this eliminates that.
You last component really is your mobo.
One symptom you seem to be having is that the os/software no longer detects the second card properly after some repeated tests. Given that the software initially detects it, then after sometime loses it, it would point to a hardware fault not software. It would be very unlikely, I would have thought, for software to detect a device and run fine, then drop a device shortly after. More likely, there is a physical problem that causes the device to no longer be visible or connected.
You could try removing one card to just run one 295, if it tests fine after a cold boot, great, restest until the problem reaapears. Then try it in the other pcie slot to see if both generate the same borked event (after cold boot, remember we want to start with it definately working)
I'm just wondering whether it is localised to a specific pcie slot.
Also have a good look with your Mark 1 eyeball at all the capacitors/chips on the board for anything obvious, you probably have done this already but I'm just covering bases here.
Otherwise, it could be chipset stuff which others have suggested but I know not much about. But again, there may be ways to try and isolate the performance/influence of individual chipsets.
Feel free to knock back anything I have said, it's hard to remeber everything you have tried/described so far as this thread is getting long, just like all your other ones.
Hope the thoughts are useful.
Tye
As to 3d program crashing when you load it, if there was no 3d device detected to run these programs, then they would crash wouldn't they? I mean Vantage and 06/05 all run a quick check of your hardware config before they finish loading, it figures that if they can't properly detect your 3d hardware then they will hang. I think this further points to physical problem detecting 3d devices.