Notch wrote a "reply" here: http://notch.tumblr.com/post/8386977075/its-a-scam
To sum it up, there's a big reason we're seeing the same objects repeating over and over, and there's good reasons why nothing is animated.
OP mentioned hitboxes, which will be a problem. I also think the lighting is subpar, compared to say http://www.gameblurb.net/wp-content/gallery/030311-unreal/unreal_tech_09.jpg
I wonder if some sort of hybrid engine would be possible. Because you'd really want voxels for static or little-moving environment, but you'd want polygons for characters and such.
Edit: Just making an argument here, but tesselation effectively already offers unlimited detail, in polygons. http://www.youtube.com/watch?v=sQQpCd_vvGU But no game has really used it to its limits yet. Why? My guess would be costs of producing those 3d models. And if you want to see something truly impressive, see how this alien changes http://www.youtube.com/watch?v=1c_PVtMIz-A&feature=relmfu