I don't think that the change to the ncr53c9x driver is the culprit, the only function that was modified isn't called at this point. I just tried a LOCKDEBUG kernel but that crashes in some of my code before this point. I will try later with a clean -current tree.
Even though this problem should've presented itself sooner, it's possible that the SCSI bus has issues. On two machines so far I've had to change my SCSI-IDE adapter settings because the newer versions of NetBSD have been more picky about the SCSI bus. In both instances, everything worked fine.
Running the SCSI bus as asynchronous is much more forgiving when you don't have proper termination and such.
John