The other thing to do is run 'systat vmstat' which will show IOPS on each disk. I find it really funny that accesses to one disk would mess up another. I wonder if there is some locking error in the code someplace. Have you tried uses /dev/rwd1d instead of wd1d? Maybe that's wrong, but I would expect the raw device.
Attachment:
signature.asc
Description: PGP signature