VM's crashing, no evidence of a hardware issue - should I rebuild libvirt.img or the Unraid USB?

[WalkerJ]

My VM's have been randomly crashing. I've done dozens of tests for two weeks and there is no evidence of hardware being to blame. Is there something in libvirt.img that could cause something like this? Should I rebuild my Unraid USB? Here are some details:

 

After lots of troubleshooting I've found that Prime95 will crash any of the VM's instantly when running the small or smallest FFT torture test. It crashes the second I click the start button and the temp never goes above 45 C. It only crashes the VM and doesn't affect Unraid. Booting the server with Windows instead of Unraid and running the same tests directly on the server rather than from a VM doesn't cause a crash, and in fact it spikes the temp to 100 C (which is the default thermal throttle temp for a 13700k, and Intel says is an OK temp for that CPU) and continues to run completely stable in spite of the temp. (The VM's were crashing randomly before. I started testing with Prime95 because it can cause them to crash on demand.)

 

I've tried everything else possible with the hardware, BIOS settings, and Unraid and VM settings. I've set the Tjmax in BIOS to 65C to prevent the CPU from going higher than that, and set max power draw to 235w, and it still crashes. There's no correlation between the temp/power and crashing. The problem is also 100% not RAM related. I've tried different RAM, one stick or two, and lowering the RAM speed to half of stock. The VM's still crash in all scenarios. I replaced the power supply with a brand new Seasonic 850w, which is more than double what the system really needs. If the problem were the CPU or motherboard (or any of the hardware) it would crash running the torture tests locally rather than through a VM because it's using more cores than I had allocated to the VM, and the temp goes much higher. 

 

There's nothing helpful in any of the logs. 

 

Here's my original post...