After an image loads, you can resize your browser text (Ctrl + / Ctrl -) to instantly reveal a different image with each resize event. By sweeping through very small and very large zoom levels, you can make peoples' fingers trace out a diagonal line from top-left to bottom-right as the registered point moves with the underlying voronoi canvas as it scales up and down.
Side note: how hard would it be to reduce the time delay between identifying a point and displaying an image? Could this be made realtime by preloading all images? The flicker would be terrible but I'd want to try it at least briefly.