I know it is hard to merge images of 3D objects, but there has to be a way to easily merge satellite images. Check this picture of the skyscrapers in Chicago and see how they are bumping into each other. From the looks of it, the images were taken either by different satellites, or by the same satellite at a different time/location.
This is an interesting problem since we want to see images of things from an angle so we get the feeling of depth. We don't want perfectly overhead shots of all the terrestrial things. Then, it wouldn't be fun to look at them. Windows live has the same problem too. Anybody who wants to allow panning a satellite image hits this problem. This issue is not noticeable in things that are not tall, and things that do not cast a long shadow. So, they could detect tall objects and try to keep them all in the same grid (same camera angle) and try to connect grids wherever they do not detect tall objects.
It will be interesting to see how this problem will be solved.