> What free speech advocates ignore here is that even they generally draw a line somewhere. Death threats, defamation, pedophilia, sharing bomb making materials etc are usually accepted by everyone to not be acceptable.
Assuming we're talking about the abstract concept of free speech (as opposed to 1A, which dictates what the US government is allowed to censor), the "line" is around expression of ideas. A death threat isn't "expressing an idea", it's coercing someone with violence. Similarly "harassment" which you didn't mention, falls out of bounds of speech because it violates another's right of association (you have the right to speak, but you can't force me to listen). Pedophilia obviously isn't an expression of an idea, although pedophilic advocacy, while repugnant, is still in bounds of free speech by definition.
Whether platforms are obliged to adhere to a "free speech" standard is a different question. Personally, I think so much of our speech is flowing through a handful of these large platforms that they are the de facto public square, and should be regulated accordingly or broken up. Even if they could articulate a clear moderation policy and enforce it fairly, simply having that much power to determine who sees what content for so many citizens is concerning. Even if you're a liberal or progressive and thus largely enjoy the alleged bias in platforms' moderation policies/enforcement, recall that in 2016-2017 we were virtually certain that Russia was manipulating Twitter to influence the US presidential election--if you believe Russia can indirectly influence our elections via Twitter, then it necessarily follows that Twitter can directly influence our elections and surely that's too much power to give to a corporation.
> 'But those are different' is usually the argument here. But why though? Because they cause harm? Doesn't inciting racial hatred cause harm?
I think the issue is that many don't trust platforms to enforce their own policies consistently. On Twitter for example, one gets the impression that it's okay to incite racial hatred toward whites, Asians, Jews, and even "ideologically diverse" blacks/etc, which is to say that the policy is neutral but the enforcement is biased--and that biased enforcement constitutes harm. Of course, a racist might respond "Good, we should punish whites, Asians, and Jews for their race because historically other whites, Asians, and Jews have enjoyed various advantages because of their race", but presumably the goal is to minimize racism.