1) While I don't know specifics, I'm sure streaming, probably oct-trees, and possibly compression/decompression of nodes in the oct-tree.
2) You need to add or modify the scene in an editor and create animations if you want those. No dynamic lighting or shadows at this time, pending hardware advances (it can be done, just not very well in realtime).
3) Laser scanning can quickly build rooms identical to real world rooms without having to go in and model everything. The flaw with it is anything behind any blocking objects won't be seen, so I suspect you end up with a partial model of a room. It is great if you want to observe the world from a single point, but most of the time you want to move around (this is my main skepticism so far). It is possible maybe with software to take several laser readings and then assemble the complete room. As I said, though, this is where I'm skeptical.
At the end of the video they say they can do animations and a demo of it will be coming soon. With them, that means around 2016+, but other companies and researchers have demonstrated voxel rendering and RAGE proves animation can be done with it.