They should absolutely try to make that set as big as possible though to increase incentives for paying for it.
Dropout's model seems to be the sanest for this kind of thing: $6/month, but for a variety of content from a cross-pollinated group of creators. It's independent and niche, but there's enough variety that something in there ends up both worth the subscription and unlikely to have existed entirely on its own.
Paying $5/month for more content, but from _just one creator_, always feels at best like a worse deal in comparison to a more expensive but broader subscription service. At worst, it turns into a gateway to Twitch-like (or OnlyFans-like, depending on the content) parasocial entitlement from fans.
The problem is that only a tiny number of users are actually going to sign up anyways, so you need to extract more value from them. Say he makes $0.50/month from the average subscriber on YouTube. If he charged $0.50/month for his Floatplane channel he would actually loose money, because the people that sign up on Floatplane are going to be above average subscribers. So he needs to charge more to break even. But as he raises the price he is even more heavily filtering for the biggest fans, which were bringing them in top percentile revenue on YouTube, making the problem even worse. This means that these platforms are always going to be priced to profit off the whales, rather than the casual users who enjoy watching some videos from these channels. Maybe in some beautiful feature where publishing on separate platforms becomes normalized this will change, but it is very far in the future and a huge roadblock to getting to that future.
My configuration, in case anyone's interested: https://github.com/geerlingguy/arm-nas
What you really need is to be able to trivially publish videos to something like IPFS, and have a large contingent of folks out there using convenient local video apps so there's a market for the content. I could imagine something like "community indexes" (perhaps a spiritual descendant of Usenet) to help find content, playlists, etc. and handle recommendation ML stuff locally.
If we had been a little more wise and avoided the walled gardens of social networks in favour of identi.ca or GNU social etc. aeons ago, we would be in a better state now in many ways.
In the early days of the internet, people would host servers out of a university lab or their home for fun or to give back to the community. This didn't scale as more people came online. There were too many people and too few hosts. Protocols moved slower than product development. People needed to download apps (and this was before app stores). So the corporate web won.
Once it gets more traffic, you get more bots, more AI tools pulling all the content, more botnets and DDoS attempts, eventually have to deal with governments and DMCA...
It's tough :(
When a sponsor is that persistent, and they offer something "free" to people, alarm bells go off.