XPS 8960, Intel RST mirrored NVMe drives — one drive intermittently disappears from BIOS

We run about 10 XPS 8960s configured the same way: two Samsung 990 Pro 2TB NVMe drives mirrored via Intel RST (the 8960 doesn't have hardware RAID, so RST is what's available). We moved the drives over from our previous-generation machines, the mirrors built without issue, and everything ran fine.

Every several months, on at least one machine, one of the two drives stops showing up in BIOS. The RAID volume drops to degraded and the system keeps running off the surviving drive.

What did NOT fix it:

• Resetting BIOS to defaults
• Reseating the missing drive in its M.2 slot
• Power cycling / pulling the plug for a while

The drive itself is fine. If we pull the "missing" drive and put it in another machine, it's recognized immediately.

What DID fix it: doing something that forces the original machine to re-enumerate its hardware more deeply than a normal POST. For us that's been either temporarily swapping in a different M.2 drive, or reseating / changing the RAM. After that, we put the original drives back in, BIOS sees both of them, the array rebuilds, and the machine runs fine for another 3–6 months until it happens again.

Best guess on cause: this looks like a quirk of the Intel VMD / RST stack on this platform. The controller seems to latch a bad state for one of the NVMe devices, and a CMOS reset isn't deep enough to clear it. Forcing a different hardware topology (different drive present, different RAM) clears whatever was stuck.

We've hit this 4 or 5 times across the fleet now and couldn't find a clean writeup of the symptom + fix. Just posting in case others run into the same issue.