Success does not require success in every niche, including extremely small high end ones. Instead, some niches are way too small for an 800lb gorilla to bother with, even if such a niche keeps a few far smaller operations like Hasselblad-Imacon, Mamiya/Cosmo and Rollei going --- just barely. Adding MF would at best add about an extra 0.1% share of the DSLR market to the 50-60% share that Canon has already. This is like saying that survival for Toyota, GM, or Ford requires them to start competing in the sector now dominated by the likes of Ferrari, Lamborghini, Bugatti or Rolls-Royce.
Well, I agree with your assessement, but, turning myself in the devil's advocate:
What has enabled Canon to move up from a second position in the film days to the first position in the digital era? Well, my view is that they did because:
- they released good products (whether they were the best or not is IMHO still opened to debate),
- by tapping into the thirst of consumers for easily understandable metric... namely pixel count and by releasing products with slighly more of that than the competition,
- by taking the higher ground with the 1ds - 1ds2.
They have sold very few 1ds + 1ds2 combined, and who can tell whether they have made money on these products. But these 2 bodies have contributed hugely to the image that Canon is the best game in town, and that has, IMHO, translated into hunderds of thousands of sales.
Just like Dell releases XPS game station that few people buy, just like AMD releases FX60 CPUs that nobody buys, just like GM produces Corvettes,...
From this standpoint, Canon must have been thinking about what it will take to maintain this image of being the best game in town.
Whether the answer is a 16MP 1ds3, a 20MP 2ds or a 33MP MF line up, I don't know, but I wouldn't find it completely absurd if Canon had also considered the MF route.
Another factor worth considering is that the 1ds2 when released cost about half the price of the MFBD with the same resolution at that time. A 33MP Canon MF body priced like a Mamiya ZD or lower would have potential for significantly expanding this market segment IMHO. This would mean more profit as well...
As far as people buying into a new set of optics etc... they did it once with the EOS line up and see the results 20 years later. A smaller company like Hassy managed to develop a completely new line up in a few years for that same market...
Finally, people feel nowadays that 16 MP is enough. This is mostly based on limitations of our current printing and display technologies. Who can swear that:
- nobody will release within 5 years new printers/papers that really show a difference between 240 and 480 DPI?
- the prints in galleries will not slowly disappear in favour of very high density TFT screens for which 480 DPI is a must? IBM released a 22 screen inch 3 years ago with a 3800*2400 resolution at 200 DPI. I saw those, they are already virtually impossible to distinguish from prints, but with a much higher contrast.
Anyway, the bottom line is that us thinking that 16MP will be enough for the foreseable future is IMHO short sighted.