I would imagine that the test code triggered by the setting of the DIP would be in ROM. I would imagine the switch either modifies the address the CPU looks at for bootstrap code, or the bootstrap code tests the state of the switch to see what section of code to execute first.
I'm mainly just guessing on that though, I suppose it's possible they could have wasted $100 worth of chips and board space on a hardwired test mode. Or it needs a special test ROM set to work.
If you want an innovative way to test your ROMs, you could try finding an old, old ISA network card, or ISA VGA card that takes a pin compatible ROM, then use something like uniflash to grab the code off them one by one, (Providing you know the address of the ROM on that card) and compare it to versions that one might *ahem* aquire by other means. If you don't catch my drift on this though, might be best to ignore it, because only a hell of a lot of explanation will (maybe) save you frying something.
regards,
Road Warrior