I have actually seen people banned from communities when coming armed with pretty detailed data and thorough argumentation. I wonder if there’s a principled way to separate low-effort trolls from those simply willing to argue the unpopular side of a controversial topic.
If I spend [not insignificant amount of time/mental energy] with you just to disprove something inane, and you do that to me n times, I won't argue with you anymore, even if the n+1th time happens to not be completely baseless. At that point it's either explicitly on purpose, or implicitly linked with a bias that makes them susceptible to bullshit that confirms said bias.
I don't know if there's any ab initio way of knowing if someone's spouting nonsense on purpose or not, so character patterns are hard to scale. But that's why moderation online is harder and faster than in real life, because you don't have a 1on1 interaction in the same way. It's up to netizens to behave and think before speaking.
Hell you can see the consequences of not holding a firm line on these sorts of things in this thread. There are people literally "just asking questions" about "men in women's sports" as an example of the "terrible reddit censorship"; as if it's not transparently obvious what they're doing.
So you have to ask yourself, even if you could reliably id anyone who wanted to argue in good faith (i'd argue this is considerably harder than you'd think though that's more of a gut feeling /shrug), if their goals are "convince the world that <race, gender, nationality, sexuality> is causing harm" what is the value of giving that a platform even if you're willing to expend the emotional energy to debunk it (over and over and over and over because even if you have some kind of metric, your users don't and if allowed someone always feeds the trolls).
Not only are these viewpoints, which you claim can only be argued in bad faith, supported by empirical evidence, but the people who dare to support them publicly are risking social and financial ruin. If "unsafe" and "marginalized" do not describe their situation, then the words have no meaning.
As a minority from one of the groups at the unflattering end of the crime rate distributions, it is my lived experience that the type of censorship you are advocating for is what makes me feel unsafe.
Inconvenient empirical truths about my community can not be discussed in the sort of polite and caring settings where the seeds of real solutions that don't ignore reality could germinate.
Instead, they can only be discussed in rude and hateful settings.
Having a 0 tolerance policy for these kinds of people makes your space safer for maginalized communities and ends up with better debates/discussions over all since, as i alluded to, these arguments aren't actually in good faith or rooted in any sort of data.