That was the only idea I could come up with. "disk unfail -s" on the partner head and let a spare take over. Although on the owned head the disks aren't failed so not sure that will work.
Could possibly fail the disk on the head that owns the disk and let a spare take over and then swap with a good disk. I think because the module failed and was replaced it set a failed state in the partner and was never cleared when the good module came online.
This is a QA system and only serves data internally. Not a prod system.