That reminds me of when my (OpenBSD) software raid failed.. I think a power delivery issue took out the first disk for a moment, leading to degraded state and a long rebuild. I panicked and ran some smart self test on the other disk. This wasn't a good idea; the test caused some commands to that disk timeout and eventually softraid figured the disk is toast so now my raid is completely down halfway through a rebuild.
I figured the disk is actually ok, it just went south because of a temporary glitch caused by me. So I read the source code of the softraid implementation, flipped a bit on disk in softraid metadata with dd, rebooted, and got my data back.