Does anybody run CDJ on an "underpowered" machine?

In my continuing quest to isolate the cause of random but inevitable access violations in CDJ, I'm curious if anyone runs CDJ on a slow machine? I've been trying to run it on a 486-66 w/16MB of RAM running Win98 as a dedicated music automation controller. I recently reformatted this machine and reinstalled Win98/CDJ (and nothing else) on it from scratch to eliminate the possibility of misconfigured software, with no better results.

I've tried lowering the serial speed at which the Slink-e talks to CDJ (I had been running at 38kbps), but that actually seems to (by a non-scientific observation) increase the frequency of traps. It makes me wonder if lost data communicating with the Slink-e is the source of the traps, and that my 486-66 is just not able to keep up. I have been able to run CDJ for 3 days straight on a 300Mhz Celeron w/64MB. I did have a trap or two on my 200Mhz PPro, but in retrospect I may have had a software configuration problem, and I don't remember the serial rate I defined when I did that. I'll probably retry that test under more controlled conditions.

I am also running with the Slink-e connected to two CD jukeboxes on two separate Slink busses. How many CD jukeboxes do most people use w/CDJ?

The only thing which says lost data due to buffer overruns may not be the problem is the traps occur at just about any time, ranging from when CDJ is actively playing CD's, to when an IR transmission is received, to complete quiescence (such as the middle of the night when CDJ is "asleep").