The tree of https://www.lesswrong.com/posts/adk5xv5Q4hjvpEhhh/meta-new-moderation-tools-and-moderation-guidelines?commentId=uaAQb6CsvJeaobXMp spans over two hundred comments from ~fifteen authors by now, so I think it is time to list the major points (and, particularly, what can be done) raised there.

Optimal moderation policy might depend on the userbase, so please take "uld" as abbreviation for "in current state of LessWrong, to proceed closer to being actually less wrong AND also build path to further success, moderation should". This post is written agnostic to whether the tactic would change later; it's an interesting discussion direction but not for this time.

So, here are the major points; when tallying them up in terms of net agreement, I was surprised (or perhaps I accidentally gerrymandered to those unintuitive results). If you read this, please agree-react on the points or their negations; I ask not to use negative agree-reacts or disagree-reacts because that would be hard to sum up.
If there is a new suggestions, I might append it to the list.

1. uld use emotional tone and user comfort as a major factor whether comment stands.
   Negation: uld not decide based on emotional tone, except borderline cases and rules stated in advance such as personal insults.
2. uld not use user comfort at all when deciding if comment stands. (at pain of writers' feedback loop that being offended or off-put by more criticizing comments will reduce amount of them)
   Negation: user comfort is an input into moderation process.
3. uld favor focus on content truth-value as the primary factor whether comment stands.
   Negation: correct comments may be eligible to be acted against, depending on their other properties.
4. uld see that comments try politely guiding author to be less wrong.
   Negation: comments may use whatever technique, symmetric or asymmetric, for the original author to not post wrong takes as frequently.
5. uld consider deleting factually wrong comments.
   Negation: wrong comments should be kept in the record.
6. uld be at least two-place, to consider whose post is the comment located under.
   Negation: uld be the same regardless of author of the original post (except, possibly, off-topic comments).
7. uld be at least two-place, and include who wrote the comment.
   Negation: uld be mostly the same for different comment authors.
8. uld maintain identical comment visibility for all users to preserve common knowledge.
   Negation: asymmetrical visibility may be implemented. (with debate participants quoting from muted-for-others comments if they deem them relevant)
9. uld allow for spaces that explicitly protect low-polished contributions.
   Negation: spaces for less polished contributions should be created with tools not involving moderation.
10. uld be designed to show who is banned from commenting on the post, without extra work. (so that absence of critic's comments is not taken as evidence for post being right)
    Negation: list of users banned from the post may be maintained separately. (to not distract from discussion)
11. uld be designed to show rebukes even if they are top-level critique posts and author of such is banned from the post.
    Negation: critical comments are under the post, critical posts are in ping-backs and author may decide not to provide them better visibility.
12. uld encourage critiques to have their own top-level post, in hopes of a more thoughtful reply, and encourage users to look through such trees of posts.
    Negation: uld discourage or be neutral on replies being separated from the start post.
13. Site features should not allow voting without viewing comments (the whole tree, or subtree of a given comment).
    Negation: site features should allow voting based on post/comment's text itself regardless of clarifications, questions and other things in comment section.
14. uld encourage voters (particularly downvoters) to write a comment explaining why they do not want more such posts on LessWrong.
    Negation: uld be neutral on users explaining their votes.