Does anyone else ever get traps in CDJ?

I can't run CDJ very long (very long = 10 hours or more) on my system before it traps. I get this on two different computers (a Pentium Pro 200Mhz and a 486-66) both running Win98. The most recent trap occurred with the 5/21 beta, but I've had this problem for several months. This most recent trap occurred while playing from a playlist, but I've had them happen while CDJ was in sleep mode doing nothing (IR traffic may have been received).

When these failures occur, typically two traps occur sequentially: the first while CDJ was playing, then the second after I noticed CDJ had trapped and I selected the button to terminate the app. The detail info from the traps:

CDJ caused an invalid page fault in
module MFC42.DLL at 0177:5f402351.
Registers:
EAX=81555a08 CS=0177 EIP=5f402351 EFLGS=00010286
EBX=5f4d1a68 SS=017f ESP=0150f918 EBP=0150f940
ECX=c15ae4c0 DS=017f ESI=5f4d1a58 FS=3897
EDX=5f417b2b ES=017f EDI=8c25a800 GS=34c7
Bytes at CS:EIP:
8b 07 89 46 0c ff 15 f4 b1 49 5f 8b 4d f4 8b c7
Stack dump:
5f4d1a68 00000040 00000022 0150f994 00000022 5f4d1a58 0150f91c 0150f99c 5f492bb1 ffffffff 0150fa1c 5f402652 00000022 0150f994 00000000 5f403ab4

CDJ caused an invalid page fault in
module MFC42.DLL at 0177:5f4012a1.
Registers:
EAX=00000000 CS=0177 EIP=5f4012a1 EFLGS=00010246
EBX=5f4d0008 SS=017f ESP=01f9faa4 EBP=01f9fad8
ECX=00000000 DS=017f ESI=00902060 FS=3797
EDX=00000003 ES=017f EDI=0069022c GS=2bf6
Bytes at CS:EIP:
8b 71 04 85 f6 57 74 26 8b 7c 24 0c 33 d2 8b c7
Stack dump:
00902060 5f405979 00000a00 0069022c 5f4d0008 00690160 5f476f9a 00000001 00690160 00690160 01f9fe28 5f492ae5 00000000 01f9fafc 5f406fbc 0069027c

When these first started happening, I attributed it to the new sleep mode added to CDJ (since that was when I would leave it running for long periods of time unattended), but have since realized it occurs in a variety of circumstances. I later thought perhaps some odd IR traffic from one of my remotes was triggering the trap, but this most recent one occurred when I'm certain there was no IR activity.

I've wondered if enough serial traffic arrives at once (from IR or the CD players connected) from the Slink-e that some buffer overflows causing the trap, but reducing the serial rate has no effect, nor changing the IR sampling rate from 100ms to 200ms.