So diffraction is a problem now? Is it an indication that lens design might be falling behind sensor design? Well, lens engineer go solve it... Is this a limitation with laws of Physics? Did they not say that microprocessor will face an inevitable limit with laws of Physics at xx nm and there is no way to go beyond? ... well, we are well into 20nm territory now and going strong....
Well Diffraction is a slight problem now, and will only become more so with more (especially denser) megapixels, that's a law of physics. Moore's "law" was never a law, it was an observation made in the '70s and just happens to have held reasonably well since. Limitations at whatever-nm process were just based on the current knowledge, and the engineers managed to overcome the problems with new discoveries.
Diffraction is different, there's no way to change the relationship between DLA and pixel size, no matter what lens/sensor designers do.
But whether it's enough of a problem is the other thing.
Canon's 120 MP APS-H sensor has a DLA of f/4. Ouch.
So there's still a good margin of MP to go before we really hit the limits, good to know. And as said, it's just going to reduce where the sweet-spot lies.
Put X lens on a 7D, you get so many lppmm at wide open, you get more at f/5.6. Put the same lens on 1Dsmk6 and you get more lppmm at wide open, but you get *less* than that at f/5.6. Changing the shape of the MTF charts is not such a bad thing, as long as you know how to use it.
In the end, the line "the lens sharpens up as you stop down" will disappear, at a few hundred MP every lens will be sharpest wide-open and will reduce with aperture size.
What it will be the worst problem for is bad lenses. Take the
Sigma 20mm f/1.8. its resolution starts low, and does get better with higher f-number, peaking at f/8 (or higher, f/8 is the end of the graph). Put that lens on a sensor with a DLA of something closer to f/4 or so and the MTF will peak at f/4 or even lower, the lens will be worse at f/8 than at f/4 for the same higher mp sensor.
So in a way it means lens designers will have to become better, they won't get the benefit of a 'sweet-spot' as mentioned.
But some other ideas I like. Like the 'split sensor in 3' to do some in-camera combining for HDR, one other idea I had is an automatic 'focus on 3 different lengths for 3 shots and combine'. OK, so people do it manually now, and it only works for static/landscapes, auto-incamera would be nice though...