<<I have read elsewhere (and I cannot remember where) that, in terms of resolution per millimetre, MF lenses are not as good as some 35mm-format lenses. I have no idea myself whether or not this is true, but the argument went that it is the size of the enlargement which puts MF ahead...>>
1. Larger format lenses do not need to have same linepair/mm resolution as 35mm lens...merely because to make the same enlargement you need proportionately lower magnfications. For 35mm enlarged to 16x20, that is 16x magnification. If its lens starts with 80 ll/mm on film, it ends up with 5 ll/mm on the final print. If you start with MF 6x6, and if you assume 64 ll/mm lens resolution, with only a 9.2x magnification for the same resultant 16x20, or 6.9 ll/mm on the final print. Similarly, for 4x5 camera lenses, only 4x magnification is required, and if the lens has 40 ll/mm resolution, it ends up needing to deliver on 10 ll/mm on the final print! So you can see that lens requirements are proportionately most demanding for smallest formats.
2. Assuming equal lens performance across the board (which we already have shown is most demanding of 35mm), there is much more film area portraying the same subject filling the same frame. Because there is more area, there is better gradation of tonality. This is best illustrated with a color photo in the newspaper, compared to the same photo in National Geographic. Same photo, much nicer rendition of the subject even though both might be reproduced to the same printed size on the page. For a fixed amount of subject, there is 2.25x as much color clouds on MF film to capture the same subject on 35mm due simply to format size. So the MF image impresses observers much more, even projected to the same size and assuming lenses could deliver the same ll/mm at the viewing magnfication. (I witnessed this more than one in actual slide shows.)
3. In published comparison tests, they always seem to take a 1st generation 35mm digital image, and compare it to a second generation MF digital image...it was SCANNED to be turned into a digital file for comparison. That is hardly an apples-to-apples test! What if the scanner is the limiting factor, and it does not scan everything that is in the original film
So while the pixel count of current dSLR might be on par with a scanned MF image, the tests do nothing to address the relative gradation of tonality seen in larger media (item 2 above).