[1] I suspect it's actually a fair bit more than half, given the semi-frequent occurrences of cases like this article that no one really wants to defend.
Also, outright revocation of 230 is just one option. Companies at Facebook/Twitter/Reddit scale would do nearly anything to keep that protection since moderation at that scale simply is not possible. This gives an inroad to all kinds of regulation.
Alternatively, your second scenario happens, and it likely leads to the destruction of large, centralized social media, and a proliferation of decentralized, hard-to-sue alternatives. This too, is a win.
If a company is discriminating on the basis of race and government officials tell them to stop, there is no constitutional problem there, even if it comes with an implied threat of enforcement or new legislation, because passing a law against discrimination on the basis of race is not a constitutional violation. The government can directly punish that behavior so there is no problem to indirectly or implicitly punish it.
If a company is allowing protected speech the government doesn't like and government officials tell them to stop, that's different. Because it would be unconstitutional for them to pass a law saying the same thing. And there are many other ways the government can punish a business, e.g. by breaking them up. So the implied threat is, censor or we'll break you up, or mess with Section 230, or some other thing you won't like. It's coercing them to do the thing the government isn't allowed to coerce them to do, under threat of adverse regulation that the same government could withdraw in response to compliance.
What the adverse regulation is doesn't matter. It doesn't even have to be specified. The problem is the direction to do something the government isn't allowed to make them do.
The constitution has nothing to say on the matter of holding them responsible for the libelous, defamatory, harassing, terrorist, etc. content they fail to moderate away every day.