I consider myself an extremely strong First Amendment advocate and "free speech absolutist" when it comes to government force, and I am also uncomfortable about where the line should be for "private entities" that manage to achieve quasi-governmental levels of power, though I do not think that any level of non-violent restriction is directly comparable to censorship via the government's monopoly on physical force.
That said, a major question for me that has been growing bigger and bigger in my mind over the last decade is what exactly should be Free Speech's answer to denial of cognition attacks (and I want to be extra clear that it is a question, I have not been able to come to at all a satisfactory universal answer myself yet). Likely everyone on HN will be familiar with the concept of resource exhaustion attacks in the context of computers and networks. DOS mitigation has regrettably become a near requirement for any significant public facing Internet service. But fundamentally resource exhaustion attacks apply equally well to humans. In the idealized world of discussion and debate and philosophy abstract models of conduct are often used, but at the end of the day "implementation" still comes down to individual humans, each of whom only has so much mental capacity and so much time. And many arguments involving reality can easily be asymmetric too: it takes very little time and energy to make an assertion but a great deal to disprove it. Reality is under no obligation to be simple and elegant to human preferences.
I have seen this get put into practice with ever greater sophistication on a lot of my favorite forums. Someone arguing BS will throw out a bunch of simple stuff that takes a great deal of careful posting to show is wrong. If people volunteer their time to answer in that thread, it does no good in preventing the exact same assertions from being tossed out again (maybe remixed) in a new thread a week or month or whatever later. Eventually people just get tired or are busy. If they try to point towards a centralized source instead, that source can then be attacked and you see "argument from authority durr" and the like thrown out. It's bad faith but in a way that exploits a lot of the norms around free speech and argumentation, which classically didn't have the same scaling and automation threats. In fact, perhaps even more pernicious is when it's not bad faith but someone who has been fed a bunch of stuff they are arguing in all earnestness, yet there simply isn't time to address them individually. They must be directed to authorities and be expected to educate themselves on it.
I'm not sure what the answer is but we do clearly have a problem as some of the previous natural barriers due to cost of speech have fallen. It is common to speak of a "Marketplace of Ideas" when it comes to Free Speech, but it's important to remember that "Free Markets" are specific real tools that require regulatory support to function and have no goals of their own. Costs need to be internalized, information and processing ability needs to be symmetric, there needs to be a shared base of legal structure and norms to work on top of. Getting too far away from that results in market failure. I don't know how that might be translated to Free Speech but I do think we're seeing a "market of ideas failure" to some extent right now, and that it could easily get much worse in the near future unchecked. There is a need for better ways to improve the signal/noise cost function, and for systems to re-internalize that the point of Free Speech, of marketplaces of ideas or markets of anything at all, is to come up with good products. They're tools to serve humanity. Bad ones shouldn't be banned by force but they shouldn't be flourishing either, if they are then something isn't working.