So to do it as pormised, game time must stand still while the renderings are downloaded. This is not a good model for a smoot game.
Not really. Look at Magic Online. (The main reason I don't trust them with online projects, but that is another story...)
Everybody has all the art of cards and avatars downloaded onto their computers, whether they own them or not. When new cards come out, you download them in a patch. Once you own the digital objects, you can use them. That way you have no problem with playing with people who own other cards and avatars. The virtual minis would basically work like the avatars. You see them if someone is there who own them and uses them. Otherwise not, since noone else can activate them - even if they have the code for them on their computers.
On the other hand, I suspect that this is one of the reasons why they are hesitant when it comes to user created content - that would cause all the problems that you described.