I am always suspicious of power supplies. In this case its for an external drive enclosure so the demands on the PSU are even more off balance than normal.

However: the smart data counts power loss / resets and that gives me no hints that anything is even remotely wrong there.

So far: Primary HBA = still under suspicion. One lane on each connector seems to have a higher error rate than the rest. I intend going to swap that out anyway. Backup HBA = definitely has a failed SFF connector - one sas lane is quite dead.

The unexpected surprise - recurring single bit errors (bit 5 always) on a memtest86. Turns out that 4 DDR4-3200 dimms are not supported on that motherboard at that speed. Downclocking to 2933 appears to fix it.

The memory situation would explain most things.

I still suspect the HBA may be on the way out. I'm going to change that because it also reduces cables/connectors. Changing the internal 8i card to 8e gets rid of internal mini-sas cables, adapters and a backplate.