The SQL in question which was problematic for me (tables renamed):

SELECT posts.*, MAX(COALESCE(comments.created_at, posts.created_at)) AS latest_timestamp FROM posts LEFT OUTER JOINS comments ON posts.id = comments.post_id GROUP BY posts.id ORDER BY latest_timestamp desc

In short, this is sorting posts by the most recent comment, with a fallback to the post date if the post has no comments on it. Hard to get rid of the grouping here and get the same data back.

I'm the OP of the DBA StackExchange post, and while yes, it does technically work in this case, you do lose some abilities. For one, its much harder to `count(*)` rows with `DISTINCT`. Also, `DISTINCT` uses a totally different mode in planning that requires first retrieving all the rows and then finally filtering them. This is _much_ slower and generally not very fun to deal with!