This is a rough jumble of thoughts explaining some of the reasoning behind the proposed policy.
Do what works is an attempt to create a firm but flexible foundation for Arbital policy, make it (hopefully) a bit self-correcting by causing people to flag bad policy, and make sure that everything ties back to what actually matters, while avoiding pitfalls of other community founding rules I've observed. It takes inspiration from Wikipedia's Ignore all rules and Eliezer's Nameless Virtue.
One important thing to note: The kind of user who badly violates guidelines of any kind tends not to be the kind of user who reads policy pages. It is pointless to attempt to guard against them with strongly worded policy %note: And harmful/discouraging to another class of user, those careful enough to check the rules. And staff checking for what tone they should use towards users.%%, instead, it's important to give whichever systems are in place to prevent obviously harmful activity freedom to act quickly, enough training to tell obvious from non-obvious cases, a playbook of well-thought-out responses to different classes of situation, and a way to easily check with others doing the same job.
Attempting to codify exactly what content should conform to seems like a fool's errand, especially since we don't know. Explicitly encouraging people to do what works (aka. create the version of the page which is good for readers) rather than pointing them at dozens of pages of guidelines they won't read should cause them to optimize for something closer to the right thing. We should, of course, offer guidelines and advice for those who genuinely want it, and link to it heavily. Just, not elevate it to required reading.
Lots of sites try to create an elaborate web of rules outlawing specific harmful patterns of behavior, and task people in positions of power (often the most productive early contributors) with the work of implementing the rules in a consistent and fair way.
No humanly readable set of rules can adequately capture and regulate the complexities of human interaction, and there are always awkward edge cases.
So there are people who are disruptive, and those with authority/responsibility to deal with it have a strict and hard to change rulebook, and complex, slow, tiring processes. This often goes badly, setting up all sorts of tensions, tying the hands of staff, causing people to be acted against too strongly and too weakly in different cases, both of which harm the community. There is a place for careful, deliberative, systems, but it should not be the first line.
By tying the community rules to "do the thing which makes more healthy community" there's significantly more freedom on the part of staff to act, and significantly more keeping them pointed at what matters (so long as the group self-polices well, which needs to be designed for socially).