Interesting.   Maybe we should be consulting some of the DEC code in the 
ancient
unix archive to see if things are being done in the right order.  It was 
working fine,
on both Ultrix and NetBSD (1.2, 1.3.2, 1.4.1, 1.4.3, 1.5.3, 1.6, 1.6L for 
those that
I have tested on this board).  I really am suspecting that it is a chip level 
thing.
But, that could be related to code usage and timing.  It was not doing 
anything
suspicious when just installing code or gently  using the machine.  But, when 
it
was doing heavy compiles and really thrashing, then that was when the problem
arose.  The particular system in use at that time was NetBSD-1.3.2.  I 
thought I
would load that up on a spare drive so I could have a drive with all levels 
of OS
up, for some testing.  I was pleased when I managed to get a 1.6L current up 
on
it, only to find that the ethernet code there was trashed, and it dropped 
lots of
packets, making ethernet transfers or installs almost hopeless.  The basic 
1.6L
ran like a champ, other than that, on the KA650 with scsi, tho.  It had to be 
hand
installed from a scsi drive brought in from a 3100 class machine.

On a similar thought, I have been running an SQ706A board in a 4000/200 for a
long time, with no problems.  Maybe there is something native to the way the
CMD boards work that is at the root of the issue.

Bob Keys