"I could certainly come up with some neat ideas to include in the camera, like automatic AFMA for near and far distances"
EXACTLY. I've been wondering for a while now why this isn't a bigger gripe from everyone shooting fast primes. Nothing from Canon about it... instead we get MA for different ends of the zoom... which is also important, but there aren't any f/1.2 or f/1.4 zooms... and accurate MA is especially important for those fast primes.
Instead we're told to just adjust MA for subject distances 25-50x the focal length.
Well, my 85/1.2L requires a +11 MA on my 5DIII body for 25x focal length subject distance. This worked great for a shoot this weekend, but it's completely off (highly repeatable) for subjects at infinity, where +0 MA works great.
Now, I don't know exactly what MA is doing in the internal workings or mathematically, but I'd imagine the camera could set 2 endpoints (near/far) & interpolate in between. That's, of course, assuming a simple linear relationship between distance & MA... which I don't know is the case.
This has also made me wonder: is it possible to find a fast (f/1.8 or lower) lens that focuses perfectly at MFD (minimum focus distance) AND infinity at a MA of 0?
In other words, when buying a lens, is it better to keep trying copies of the lenses until you find one that works with your body at MA +0, b/c then it'll focus perfectly both at MFD & infinity? Or is there always this discrepancy between focus at MFD & infinity in these AF systems that don't use a constant feedback loop + phase detection system to get optimum focus (but rather use phase detection data to calculate an offset to move the focus element).
At the heart of these questions is the question: how does the MA offset work? The subject distance scale on the lens is not linear... it's more logarithmic (you have to rotate the lens more to change focus distance at closer subject distances). So is the MA offset just a certain distance offset for the focus element that the camera adds after determining how much to move the focus element from the phase detection data? If so, such a focus-element-distance-offset would have to be corrected for depending on subject distance... since you have to move the focus element less & less to change focus at further & further subject distances...