See for example the specs collected at:

http://www.open-content.net/specs/

So, a random small fry with a popular big file could put it on their underpowered, underprovisioned HTTP server -- but then as it grows in popularity various other HTTP mirrors, including end-user P2P caches, would pick up the load.

It's not exactly the same thing, but the Coral HTTP-based content-distribution-network (CDN) is also interesting:

http://www.scs.cs.nyu.edu/coral/