you have a process time limit, this is why this is happening. When a process
exceeds its limit, the kernel is sending it signal 24 for which the default
action is to terminate. You can catch it and exit gracefully  but generally
if you get an XCPU it means that if you dont get out of the way soon youll
be more explicitly killed (with a KILL). You can disable the time limit 
if you wish. 
You can see the limits with ulimit.

However, because lower down you mention only 20 megs of swap, 
you _might_ want to verify the above with a simple cpu hog program that
catches signal 24 to see if this really is whats happening. If the system
is nit XCPU'ing them, then somethign else might be wrong. 

hope it might be of help
gregg

On Tue, 6 Oct 1998, Ronny Svedman wrote:

> 
> Hello =)
> 
> I have happily been using NetBSD 1.3.2 on a Microvax II for about two
> months now, but still there is one thiong that annoys me.
> 
> when i load the machine really hard (about 5 or 6 according to systat),
> running a lot of processor intensive processes as a normal luser,
> sometimes processes terminate without explanation. This gets particularly
> annoying  if you lose your "screen" session, beacuse that kills about
> everything you are using.
> 
> I tend to put rather heavy load on the machine since i run a lot of things
> in parallel, to save surveillance time (i dont like sitting around waiting
> for compiles, file transfers, heavy bignum calculations, benchmarks et
> al)
> 
> so i run a screen session on the console (vt320 - lovely gadget ;-) ) with
> screens for three ssh shells to our club server, two ircII clients (ircnet
> & efnet), one bash shell and a su bash shell, and a systat screen (just
> 'coz its kewl- as i would say if I was still an Atari ST Warezkid
> *shudder*)
> 
> things usually work fine doing this, but every now and then processes
> start to die, and sometimes the screen process goes first.
> 
> 
> My question is: is tehre a limit as to how much one luser can load up
> processor- , memory- , and processwise ? 
> 
> how do i in that case remove this limit for a specific user?
> 
> (the machine is at 0% idle for hours at a time, especially whn compiling
> stuffs while doing other things)
> 
> the OS itself is rock steady as far as i have seen, the only trouble is
> that i have to login single user and do manual fsck's after powerfailures
> (i have tender fuses - i cant vacuum and compute at the same time ), since
> the kernel panics during boot otherwise - but this is a known problem tat
> has been fixed, if i remember correctly.
> 
> Oh, and there IS enough swap space, according to systat. (about 20 % used
> of the 20 megs available).
> 
> any hints ?
> 
> the machine has 9 meg ram, delqa, dhv11, one rd54 and two rd53 on a
> controller (digital, not sure of the model number offhand - rqdx3
> perhaps?) and a tk50 with controller. all in a BA123.