With typical torrent use, you have to wait the entire download time before you can watch anything because the pieces are randomly selected. But adding streaming just means you have to bias toward the early pieces just enough to have a decent buffer. After that, you can still be random.
The case of prime time seems to be better for this approach, not worse. People still won't all start at the exact same time. The early arrivals will all replicate the early pieces, beefing up the ability of the swarm to get the pieces everybody wants.
I haven't looked at the code, but it seems to me that as long as it's flexible about the amount of time it takes to fill its initial buffer, and as long as it keeps serving after people are done watching (to compensate for the up/down pipe asymmetry), then the swarm would survive just fine.
Indeed, I think making swarm participation much easier and more appealing might increase the depth of resources around any given torrent, making results net better for popular files.
The rarest piece is picked (from the client’s point of view of the swarm)
If two or more pieces have the same rarity, pick one of them at random
If you're shifting to a demand-driven system, then pieces aren't needed equally. A numerically even piece distribution isn't what the swarm most needs. Indeed, I start watching things on Netflix that I never finish, so having more people with early pieces of something would better mirror actual demand for them.
I'm sure that plenty of people in the Bittorrent community are worried about change, but plenty of people in every community are worried about change. Until I see some math or some simulations demonstrating that this can't work, I won't be persuaded that there's an actual problem beyond the (reasonable and legitimate) fear of change.
Second, seeding in enforced during playback, to "give back".
Third, there is a time randomization in which hot spots (pieces who are big demand) are also the ones who are the most shared (since everybody is seeding these).