I was reacting to William's error message:

  wd2a: device timeout reading fsbn 268435392 of 268435392-268435519 (wd2 bn 268435455; cn 266305 tn 0 sn 15), retrying
  pdcsata0:0:0: lost interrupt type: ata tc_bcount: 65536 tc_skip: 0

That transfer is FFFFFC0 to 1000003F, so it looks like it crosses the
boundary.  Does the code switch to LBA48 for the whole transfer?  Does
it need to be split into a non-LBA48 and a LB48 part?  Perhaps the
drive can't handle such crossings.  I could be off by some power of
two, though.

  As both drive works fine if hey're used alone, I'd suspect power supply issue:
  the PSU may not be strong enough to drive both disk when active 
  at the same time.

I concur that this could be the problem; I've found RAID provoking RAM
errors (I think).

With raidframe, everything is offset by the 63 blocks that the raid
partition starts, plus 64 blocks (guessing that the setup is normal).
So RAIDframe could provoke dd or other operations that are normally in
round numbers and wouldn't span the boundary into doing so.

Also, there could be more than one thing wrong.