Especially that most new platforms completely agree with you (ie. they don't want to allow arbitrary user content, because moderation is hard, spam/scammers/4chan-raids are bad for the user experience, and selling ads is hard if your platform is constantly in the news as a great new thing with the thorny little problem of a small nazi community in its side).

But those who benefit from the endless firehose of low-quality user generated content (FB, Twitter, TikTok, YT and so on, mostly the old guard) have a vested interest in putting the cost of the externalities on society, and free speech related concepts serve as great explanations and arguments.

Obviously we know the real cost of censorship and the positive value of an effective 4th branch.

And since there is some relationship between moderation, webhosting, DDoS protection, centralization, monopolies, regulatory capture and the previous "censorship bad, speech good" intuition. So it's not surprising that the latter always ends up overshadowing the whole concrete object level question.

The solution seems to be an unsatisfying acknowledgement that as long as the threat of authoritarianism, populism and their ilk are real the intuition to push back against state mandated moderation seems to be the correct one while also pushing back against the groups that support the aforementioned ideologies.

So as a consequence it would be great to develop a better model than the current seesaw of initially pushing for the intuition then doing a sudden 180 degree reversal and trying to scorch the earth of even the memory of some website.

Twitter has timeouts (and now AI speech police). Stock exchanges have circuit breakers. Do these make sense for Cloudflare? Maybe, maybe not.