There seems to be some consensus among some photographers that 12 MP is essentially enough. I don't really share this opinion, but mention it for a reason. I made a few comparisons between my APS-C camera with 12MP and full frame at 24. The differences in file were large but very small in print. In one case the prints were impossible to tell apart, although the files themselves were a world apart.
The other issue is that sharpening matters a lot. The image below is from an analysis I tried to do from images from Imaging Review, comparing the than new Pentax 645D with the Nikon D3X. The images are 300 PPI scans from crops corresponding to A0 prints. The images are from left Nikon D3X (landscape sharpening in LR), Nikon D3X (deconvolution sharpening) and Pentax 645D landscape sharpening. In my view the deconvoluted Nikon D3X image was sharpest in print. On the other hand, the Pentax 645D could also be more intensively sharpened. The reason I tested deconvolution on Nikon was to compensate for OLP-filtering.
For full size: http://echophoto.dnsalias.net/ekr/images/Pentax645D/A0_print_center.jpg
The third point that my experience is that my Pentax 67 doesn't 'beat' my Sony Alpha 900. There are areas where more resolution may be tweeked out from a Velvia or Ektar 100, but in general the image quality I get from the Alpha is superior. That said, it is possible that the Pentax 67 enlarges better, but at this stage it is not very clear to me that so would be the case.
Now, we use different equipment. Tim has tested the Mamiya 7, which has very fine optics while the Pentax 67 90/2.8 is probable more like "adequate" than "excellent". My scans were mostly with a CCD-scanner at 3200 PPI (Minolta Diamge Scan Multi Pro) but I also had two 6000 PPI scans made by Dominique Ventzke (who was most helpful in every way).
A final point is that my experience is that with digital I just set up the camera and "nail" the shot. With analogue I need to do focus bracketing and scan the image to find out which to use. A microscope would help, but a 15X loupe is of little help. To this comes turnaround time from the lab, in my case > one week.
An additional observation is that Bernard actually has a good point on stitching. Many subjects are not well suited to stitching, waterfalls come to mind. On the other hand, todays stitchers are pretty good, and we can buy an automated pano head for less money than what we would pay for a decent lens.
The two articles below compare the Pentax 67 with Sony Alpha 900.http://echophoto.dnsalias.net/ekr/index.php/photoarticles/59-sony-alpha-900-vs-67-analogue-round-2http://echophoto.dnsalias.net/ekr/index.php/photoarticles/16-pentax67velvia-vs-sony-alpha-900
I'd also like to add that I much appreciate Tim's article, and I don't argue with any of his findings.
From my experience running the test, you can blow up a 24Mp DSLR with a good lens to 20x24 and people will accept it although in a side by side with a higher res camera it will show (just). I imagine the Nex7 will be close but possibly 16x20 may be the limit. (I'll be testing this when I get rid of my 5Dmk2)
MF film is happy up to just over 20x24 and will look good up to 30x40 although grain will be showing unless treated and then it could look a bit 'plasticky' depending on NR
MFDB should be OK to 30x40 for landscapes although with Portraits I imagine you could get larger with acceptable results
LF 4x5 should be good for 50x40 or maybe larger if you accept some noise.
8x10 you could get to 60x80 and show no grain..
6x17 is interesting, I've found this produces very sharp 120's that easily exceed 4000dpi (with good technique of course) and you should get a 25x70 print from a 4000dpi scan.