So whether you use a lens with higher resolution or upgrade to a sensor with higher resolution, your total system resolution should show improvement (and PERHAPS, but not necessarily, also MTF50, which is the best predictor of perceived image sharpness; this is a change from my original wording quoted in the comments below), albeit by ever smaller (and less perceivable) amounts. Please note that more pixel-dense sensors will yield softer images at the pixel level, but with the same lens and focal length will record more subject detail.
BTW, I believe most good, sharp lenses have maximum center resolutions of something like 70 to 100 lpm (read "line pairs per mm," also lp/mm), but I'm not sure of a source for those numbers. However, Zeiss.dethis page at Zeiss.de talks about theoretical resolution limits of lenses. Also, the top line on this graph is for the ultra-fine-grain Technical Pan film used in William Castleman's resolution tests like his 70-200/2.8 shootout. TP's resolution value was obtained from this other Zeiss.de page. The other curves' Rs values are from this handy photo calculator and from this D2X review, which both give the theoretical max resolutions of these DSLRs as calculated from the respective sensor and pixel sizes.
It should also be noted that once the DSLRs' low-pass/anti-aliasing/anti-moiré filters are taken into account, their values for sensor resolution drop a bit, maybe around 10-20%, which will vary camera to camera, but I don't actually know how this effect can be quantified precisely. You may recall the full-frame Kodak 14n aka Pro SLR/c and the new 1.4x-crop Leica DMR digital back have no such filter in front of the sensor and thus deliver much closer to their theoretical sensor resolutions, albeit with distracting moiré under certain conditions.