However, your implication that the only real need for compositing relates to dynamic range is misleading. There are many other reasons to composite multiple exposures where you need exact registration.
Hi David,
I agree, there is a number of other applications for layering multiple images, and they might be relatively (and I know that things may be more complex than meets the eye) easy to implement.
One can e.g. think of (averaged) Darkframe subtraction for long exposures, which is relatively easy to implement because registration is a given. One can also think of averaging, like in Median averaging to remove moving subjects from a stationary scene, or multiple exposure averaging for noise reduction (although exposure stacking is more efficient but not always feasible due to shutterspeed limitations), or super-resolution enhancing drizzling (or support for multi-shot sensors).
One of the hurdles is (sub-pixel) registration accuracy, so it's more involved than just layering some images. But the fact that (as I understand it) the raw data is kept Raw quite long in the processing pipeline, in linear gamma space, opens up a lot of possibilities, also for adding real deconvolution sharpening layers which can be applied locally.
Cheers,
Bart