As someone who has studied P2P, none of this make sense to me. If I remember correctly, the Bittorrent protocol tries to maintain only a handful of connections

As someone who has studied P2P, none of this make sense to me. If I remember correctly, the Bittorrent protocol tries to maintain only a handful of connections at any given time. Parallel browser downloaders are capable of opening more connections. Also, I don't see how the discovery process would stress a network. I would expect the actual download to place much more load on the server. For instance, I could be wrong, but I think a bittorrent client only needs to contact a tracker to discover other peers. Every iteration after, the client will try exactly one new peer. That hardly sounds stressful at all! The last point you make is the only one I agree with. You've offered your users more theoretical bandwidth than you can actually deliver, which is common practice. Whether they are trying to use that bandwidth watching youtube videos or downloading with a P2P client, they will be using more bandwidth than you can handle.