That's the only way this can scale for regular users, i think. And of course it may use social mechanisms like feedback "you're the reason why this rare work of art if available for everyone".
As for scalability of TCP connections , you're probably right, not my area of expertise , but once this will be declared as an important BT problem with great practical value, i'm sure some researchers would be happy to contribute and might solve this.
Whether this can be retrofitted into Bittorrent not clear. Conceivably one could write a bridge between IPFS and Bittorrent such that an IPFS url could serve as an "IPFS seed" much like a HTTP url can serve as a BEP 19 webseed.
1. At torrent delete , set a flag on torrent to hide
2. At torrent display , don't display hidden torrents
3. Periodically or on storage_full event - fully delete hidden torrents , based on some simple strategy like "remove those with most seeds" ?
And after building it and testing a bit , creating some nice web page , a bit of marketing , and gather developers/users ? Who knows, might become popular .
Might have tried to do so myself , but i'm not really a developer.
Anyway, is it realistic?
(It would also need to have good upstream-bandwidth management, use a number of TCP connections that's less than O(n) in the number of torrents, and offer a few bulk operations like "stop seeding torrents with more than N seeds" and "resume seeding torrents with fewer than N seeds".)