On an 1:10 asymmetric connection, users would need to remain available for seeding 10 times longer than it takes to download (and presumably watch) the film to have a net positive impact on the system.
That's not exactly realistic, considering that a lot of traffic goes to mobile devices where this would outright kill battery life.
Bittorrent can be damn quick. Haven’t used it in a while but it was certainly fast on well seeded items.
With enough bandwidth Bittorrent would be less efficient than a single stream TCP between the source and destination because of the protocol overhead.
On a good day, that works out to about 100/2, thanks to my provider vastly overselling their shitty overcongested network as "fiber to the home". (It's just cable.)
In eyeball networks the bottleneck is not upload. It is traffic being brought into the edge network from upstreams (netflix/photos/etc)
Most Comcast plans in my area come with 5Mbit/s upload, even if the download speed is 15x to 25x that.
There's absolutely no reason you can't have symmetric cable if cable companies wanted to pay for the infrastructure.
Yes, you can configure it to use more channels for upstream, but only if you cut off television to all of those paying customers, likely losing a lot of subscribers, breaking carrier agreement contracts, lose the advertising dollars, and go out of business before you even get there. You also have to swap out all of their cable boxes, and leave whoever purchased their own box hanging.
As far as I know, no ISP in North America is currently using any DOCSIS 3.1 features for upload, and they're barely taking advantage of the features for download. Moving to OFDMA is the big barrier to opening upload speeds. The protocol supports it on paper, but turning paper into massive co-existing television and ISP infrastructure is not trivial.