Then you can ignore that, there's also fs corruption on disk4 that I missed when I first looked since log is heavily spammed with the oom errors.
Mar 17 12:18:05 unRAID kernel: XFS (md4): Metadata corruption detected at xfs_dinode_verify+0xa5/0x52e [xfs], inode 0x90d66a8e dinode
Mar 17 12:18:05 unRAID kernel: XFS (md4): Unmount and run xfs_repair
Many using that including myself, post a screenshot of the network settings for both servers, also you'll need to use the 10GbE IP address from all transfers you want using it.
I've seen some logged errors recently with LSI and spin downs/spin ups, but they appear harmless, not actual errors, and never disable a disk, but you could try disabling spin down for a couple of days and see it it makes any difference.
Logs are lost after a reboot, so no way to see what might have caused that, you can try this to see if it catches something in case it happens again, but it might not if it's a hardware problem.
I suspect the problem is controller related, and I would recommend replacing that rockeraid with a recommended controller.
If you unassing disk3 and start the array the emulated disk should mount correctly, then you'll need to rebuild, ideally on a new controller.
Disk16 should be clear but since I don't want to assume anything in this case, re-assign disk16 and rebuild it (it will still show unmountable), when the rebuild finishes unassign disk15 and start the array, then post new diags.