What I think the next step will be in compact digital cameras:

When you press the shutter button, it's not a single exposure. The camera samples the entire full-rez CCD at a very high frame rate, much like a tiny little movie, then, internally, combines the images in an intelligent way which uses the temporal sampling to remove the grainy noise and also improves the sharpness of objects which are not moving within that frame. For objects which are moving within a frame, it locates the least blurry sample and uses that.

In that kind of a situation, the small amounts of hand shake that you get from a handheld camera would actually improve the image instead of messing it up. It's possible to get sub-pixel resolution and come out with a sharper image if you can intelligently postprocess multiple frames taken from ever-so-slightly different times and positions. It's a technique sometimes used to improve the resolutions of planetary surface photographs from space probes, for example.

Of course this would only work for certain kinds of photos, you'd want to turn off that mode for action shots (or have the camera intelligently decide when to turn off the mode).