Then they shoot with motion control rigs ONLY. No steadicam, nothing. There's no camera operator in the world that can do the exact same move twice.
I don't think this can be true (at least not the way Lindelof describes it), because a shoot with 3D cameras would then be a whole less troublesome and complicated.
That's true, unless it's motion control, you really could not get the exact same move. I'd assume that they're doing it as sort of a reference pass to assist in the conversion. They won't be using the 3d footage directly, but as a reference to line up their 3d cards when they do the actual conversion.