I was working on some huge images from NASA a while a go and read about the distributed pixel cache. Pretty exciting, though I didn't need it at the time.
However, it gave me an idea which I think comes natural to almost any developer, namely this: If you're using the distributed pixel cache, and some pixels are already on another computer, why not have that computer operate locally on the pixels already there? An instance of ImageMagick is already running to handle the cache, and a lot of image operations are "trivially parallelizable".
This idea is not mindblowingly difficult to come up with so I assume you've already considered it. Is it something that you think will eventually get added?