I wonder whether or not the brand and legal credibility of e.g. YouTube will give them enough leverage to someday "crowdsource" the infrastructure with similar tech, but not defederate, leaving data accessible to the users. Or whether they will need to go the WhatsApp route of closing a federated system(YouTube would do this through drm and an encrypted protocol I guess).

A problem with WebTorrent is that there isn't much incentive for regular clients to support it, while WebTorrent clients have a lot to gain.

In general, it seems like WebTorrent clients will be more inclined to "hit and run" (i.e. download some content, then navigate away without contributing back to the swarm), leading the worse swarm health.

That's an issue with regular bittorrent too, yet it's still there, even when most of the users have asimmetric bandwidth.

A good use case for webtorrent is streaming, where the user will keep the webpage open for enough time to contribute at least few megabytes.

I agree. Only thing I've ever used webtorrent for is quickly sending files to other people, which it excels at. Much easier than properly setting up a torrent, stuff like instant.io is really easy for other people to work with too.

It could be that just by browsing your website, your visitors start seeding from localStorage/IndexedDB. Most clones only use memory but you can certainly persist chunks if you'd like to.

Does anybody know if there have been attempts to use this browser based P2P tech to create a decentralized Kat.cr replacement?

Zeronet has a couple sites, and there's a kat mirror there too. Trouble is in still needing admins, which zernet doesn't do nicely.