I saw a 'Kinect view' somewhere which had three players being scanned into the application at once.
In any case I think its important to seperate the practical number of players which can fit into the frame of view from the technical side. Because the more active the player is in a game the more space needed and hence the fewer number of players which can fit into the view of the cameras. More active = more technically demanding coinciding with fewer players being able to fit which mitigates most of the issues with regards to number of players in an active sense. If a game is more passive, lets say for instance its a quiz game like on the E3 2009 teaser they could easily 'fit' 4 players into the view and each player is less active so each require far less processing.
Tease.
