Need help explaining crashes 4.3BSD on Vax750

[dem at uwslh.UUCP]

I need some help/ideas about the cause of series of crashes.
We are running 4.3 BSD unix on a VAX-11/750 with 8 M of memory.
Recently, in an attempt to make better use of memory I modified several
constants in the kernel.  These are:
CBLOCK set to 256 - The size of a clist block.
		I also decreased the total number of clists since they
		are so much bigger.
NBUF set to 366 - the number of disk cache buffers
BUFPAGES set to 1024 - the number of 1024 byte disk cache buffer pages
I ran config, make depend, and make vmunix.
The new unix does appear to have better performance, but it crashes much
too often (3 times in 2 hours the first time I tried it, but not till it
had been up for 1 1/2 days and twice in a day the second time, but
not till it had been up for 2 days).
The crash is very strange.  The system does not go through the usual
panic and dump, but simply halts with the codes:
800011f3 04
and then reboots.  The address is in something called Xtransflt and the
code 04 means interrupt stack not valid or unable to read SCB.
The first time I saw this crash I figured it was a hardware glitch,
but after several of them we backed up to the previous unix and the
problem went away for a week.
We made a newer unix and the problem repeated.
Does anyone have any idea what is going on?
Is it a hardware problem (possibly brought out by the software) or have
I broken something in the kernel?
One other item which may be relevant:
Since switching to 4.3 we occasionally see crashes with the message:
tbuf par fault
with codes indicating cache error.  We did not see these under 4.2,
however we also had our vax revised to level 7 shortly before
converting to 4.3 BSD.
I would appreciate any help or clues that anyone can provide.
-- 
David E. Miran         ...!{seismo,harvard,topaz,ihnp4}!uwvax!uwslh!dem
Wisconsin State Hygiene Lab    or    uwslh!dem at rsch.wisc.edu
University of Wisconsin          (608) 262-0019
465 Henry Mall
Madison, WI  53706