I guess that once we reach the point where a sensor can have very high read out speeds and processing to match, there's no reason why you couldn't take a photo at say 1/20th of a second, but it's actually made up of a few hundred or a few thousand separate images, all at a very fast shutter speed and crazy high iso. Then the camera can stack them all together into one - stacking will bring the noise back down to controllable levels, and the individual images could be aligned to eliminate either camera movement or subject blur. Or save both processed images, or even the massive raw file of the lot for you to sort out later should there be a few things moving in different directions.
Something tells me we're more than a few years off having the technology to do that, but it should be possible eventually. In the meantime we can carry on using IS and panning.