Having a completely decentralized solution also comes with the issue of future governance. If a single entity controls the direction (even if the spec is open and you can host it yourself), then it's not decentralized. If you end up with a consortium then you'll face the same issue of email, innovation is hard to spread as you need multiple actors with competing interests to agree.
If your vision is having multiple entities providing different experiences tailored to individual taste, they might start consolidating and effectively forming several disjoint platforms.
p.s.
The web can be said to be decentralized but it's dominated by large players all the way from hosting to browsers. If all three major browsers don't agree on your proposal, it's effectively dead. Who's to say entrenched players won't arise in your vision of a decentralized social media?
Email for example can be thought of as a social networking app but it's really decentralised.
While you can ban Gmail, it's really hard to ban Email.
Something like AT Protocol would be what it would like like or activity pub.
But so far, they are all so bad.
Email is a bit of an outlier because it gained critical mass before the web was predominantly commercialized.
So a fountain of child sexual assault material?
We all walked into the walled gardens and went "ooh, looks mighty nice in here!"