Hi again,...
The issue still exists an until now I (or Tyan or Hitachi or Nvidia)
hasn't found any solution.

Could you please tell me if the sata_nv driver uses Native Command
Queueing (NCQ)?
This could perhaps help me to further trace the problem down, although
this is not very likely as the issue appeard on IDE disks, too, which
shouldn't use NCQ ;)

Thanks,
Chris.