May I timidly dip my oar in this murky water?
Jack's observations drawn from comparing prints speak for themselves and they are exciting for FF dDLR shooters. But the detail that any digital sensor can capture depends on the pixlel pitch, and the aliasing filter. Ultimately, the result has to be in accord with the sampling theorem. Looking at the print, as in Jack's study, does not shed much light on the situation - i.e. it is hard to extrapolate those observations to other lenses and imaging situations. Assuming that the lens is the weakest link in the chain for 4x5, the question is how much detail will be recorded on the film that is above the noise threshold. Under some situations it is plausible that the 1DsMKII could look better on the print, but i don't think we can assume that will be the case for all MF lenses and imaging situations. The most objective test I can think of would be to choose a lens and try to shoot a line source and compute the MTF after raw coversion. Then we could see how limiting the 1DsMkII sensor is.
On the other hand, just because the scanned 4x5 has more pixels in the image does not ensure that the resolution is higher. Once you have enough pixels to capture the info passed by the lens and satisfy the sampling theorem, adding more does not gain much.
This is a complicated business.
Just my $0.02.