I apologize, I hit the "send" button too early. Here is what I wanted to say.
users with tension between them are another topic entirely
That's the second thing I've described as a mistake above: the illusion that there is an issue of offtopicness on the MLs that deserves moderation in the sense of makin it justifies to turn to a warn then mute approach. There is arguably none. What there arguably is instead: - offtopicsness, which never makes it justified to turn to anything like "a warn then mute approach", and is simply handled by participants, by not responding to this or that message; - behavioural / relation issues, which might justify turning to the "warn then ban approach" when all other failsafes have failed, but are often more fruitfully handled using the "pull them aside and help them iron things out" approach I am advocating.