For the sake of curiosity, I have butchered my card a little further. Unhappy with not having a definitive answer about the SIMM organisation issue, I cut the RAS# traces leading to the SIMM socket, re-connected the lines for the first rank/"side" of the module with bodge wires, and used pull-up resistors for the other. This should have the effect of forcing the second rank to remain mostly idle, never dumping anything onto the memory bus.
The result is that the same amount of memory is detected, and the SIMMs generally seem to remain somewhat cooler. As far as I can tell, we lose nothing, as the InterWave doesn't directly provide a way to make use of dual-rank SIMMs (no capacity is lost).
One thing I have noticed since disabling PnP-compliant mode is that the Gravis setup utility will not run, claiming that it can't find the card. This could be because it detects a PnP-compliant BIOS, so maybe the problem will disappear on older machines. It could also be that I need to revise the circuit used to disable PnP support, which I will try when I have time. The important thing will be to get feedback from someone with an actual non-PnP system, since the official system requirements call for a 386 as the minimum, and there are no plug-and-play 386 machines, as far as I know.
EDIT: I just tried disabling PnP on a genuine Gravis card (by fitting JP9), and the behaviour is the same on my system; SETUP.EXE refuses to run.