The cameras fimrware breaks down the image-receptor into a 4x4 grid of 16 rectangles, treating each of these 16 rectangles as it's own seperate sensor. There is a monitoring feature that measures the number of photons that hits each of these 16 rectangles. The camera can then turn off the rectangles individualy when the saturation number of photons is reached. There would be some sort of stitching routine that would smooth out the edges of the rectangles.
What good would that be? No more blown skys. Boost in dynamic range.
With a 45 point focusing system, how much trouble would it be to have a 16 point metering system?
What do you dream of seeing?