>>1. This problem didn't manifest itself until Novell 4.x/Client 32.
>>2. AFAIK, it doesn't occur under NT.
>>3. I haven't seen it occur on machines not running Client 32.
>>4. This only occurs after the machine has been running for sometime and occurs sporadically.
>
>It might also be worthwhile to measure the ethernet utilization when the problem does/does not occur. It might be a swamped network problem, and that the client32 machines are not tolerant of the errors. (We have DOS client32 machines that visually report a "can't find server/retry" message to the user. The managers think the machine is bad. The non-client32 machines have the same problem, but since they have no visual prompt, they just get stuck waiting for the server for a half-second and resume.)
>
>We've also had high-utilization situations kill the MS client for netware (which had a 20% resource overhead) After changing to the Netware Client 3.01, the frequent unexplained crashes stopped.
>
>I guess the best test would be to run a version of the program with local data. If it doesn't kill a stand-alone machine, the problem is somehow network related- either server, site, or client.

Well, utilization wasn't an issue in my case. Less than half the available connections were in use, and a relatively low number of users on my stuff. It pretty much occurs the way described. The circumstances under which it occurs, however, makes me think that its either directly related to the Network OS or other programs that are started during a session. Your suggestion of a test, however, is a good one.