Is the second 3c589 you tried exactly the same model as the first, or slightly different, eg. does it give the same output with card_info?
I wonder could it be because of the 68000, but I'd expect an odd address access to give a guru rather than a freeze.
It might be better to test the card with sanautil from Aminet instead of a TCP/IP stack:
sanautil device 3c589.device status
