I think sorting through comments like that would be an excellent application of AI. I know some sites use AI for that in a general way. But you could train it yourself to recognize toxic stuff yet still know when someone who, for example, was using foul language but still saying something useful or making an good argument.
It'd be like having a friend that could look them over and say, "Oh, you don't want to read that one. Or "This guy's a dick, but he's got a good point."
Except your friend wouldn't have to read them either, the AI could do it.