Just doing a coredump of the information I have gleaned.

My Netware 6.5 SP4a has had the following services disabled:

CIFS unloaded at behest of Novell Engineering
NWAGENT unloaded by me since any backup attempt hangs the box
NDPSM unloaded by me to prevent ABENDS from NDPSM
BROKER unloaded by me since NDPSM is disabled.

I also tried setting the U320 drives to U160 speed at behest of Adaptec to
attempt finding a problem on the SCSI bus itself. No effect. I also have
the latest firmware and drivers...No effect.

These tweaks resulted in a box that would crash randomly every 3 or 4 days.

I was asked by Engineering to monitor busiest threads during a backup.
I did so and FSTAPE2 would bounce around near the top of Busiest threads,
and at one point became the single busiest, then fell back to around 3rd or
4th. The kicker came when FSTAPE2 suddenly vanished from the list. I
assumed at that point an ABEND had occured so I toggled out to the shell
and sure enough the classic SYS and USER volumes have been dismounted due
to an I/O error in AACRAID has occurred. Was displayed, NSS Pool corruption
blah blah. ABEND.LOG reports CPU hog with AACRAID.

Now, one may be inclined to think there is something wrong with the NWAGENT
or AACRAID. I think the problem is a little more insidious.

What I have found is that if I disable processor 2, the problems seem to
vanish. I decided to disable processor 2 on a hunch since I was seeing
lots of reports about the need to disable "HyperThreading". Since my box
is a dual Opteron of course that had no meaning. but i was thinking maybe
all of the reports about disabling "HyperThreading" may be masking an
inherent flaw in multiprocessing with SP4a.

The system has been up now for 7 days which is a new record.

My next test will be to enable NWAGENT again and see if my backup hangs
disappear. If they do I will then move on to re-enabling NDPS.