This is interesting, I've never considered torrents for this exact case before. Has anyone done any groundwork to figure out what this would look like from a systems level? I'm having a hard time wrapping my head around the big picture - where the seeders come from, what are the incentives to keep certain kinds of data resilient, how to keep complexity away from the clients, etc.
Memento mori.