Has anyone noticed how it is sometimes necessary to flush the route table (and the arp table) multiple times, especially when switching back and forth between interfaces and/or networks? I just got done configuring a pair of LinkSys WRT54G's with Alchemy (the almost-opensource firmware available for them, which supports bridging mode, among other things - including boosting signal from the default 23mw to 84mw) - which I accomplished, BTW, entirely under OS/2 with Firefox (as if there would be any doubt!) - and in the process, had to switch back and forth between interfaces and also had to reset one of the WRT54G's about a half dozen times, as I managed to lock it up. Every reset, of course, meant a return to the default subnet of 192.168.1.0, which was not the subnet I needed to configure, so I had to reconfigure my NIC each time.

Anyway, it appeared that after several reconfigs, I couldn't even ping my own interface, but the machine across the bridge could get a response from it (same subnet). Quite weird.