strange output from ps when mt hangs

[Andrew Dalgleish <andrewd_(_at_)_axonet_(_dot_)_com_(_dot_)_au>]

I am having some difficulty with my scsi card+dat combination.

I am running 2.8-current, GENERIC with RAIDframe.
(Some of) the relevent bits from dmesg are:
aic0 at isa0 port 0x340/32 irq 11
scsibus1 at aic0: 8 targets
st0 at scsibus1 targ 0 lun 0: <ARCHIVE, Python 04687-XXX, 6580> SCSI2
1/sequential removable
st0: drive empty

If I try to access the DAT in anyway, mt hangs.
The only way I have found to remove the hung process is to reboot.
I can not kill mt, although I can kill the parent shell.

What really piques my curiosity is the "DW" flags from ps:

18582 p0- DW      0:00.05 mt status 

given the relevent bits from man ps
             D       Marks a process in disk (or other short term,
uninterruptible) wait.
             W       The process is swapped out.

I would have expected that any process which is in a short wait would
not get swapped out.
Has anyone seen this before?

Regards,
Andrew Dalgleish