Oh, in the thread Jun
gave above,
Don and Jun were talking about the sorting needs and the performance
implications of the tree_sortkey approach, so by "sorted out" Jun must
have meant doing some tests and deciding when it makes sense to use
tree_sortkey and when some other approach would be better. And I
guess the question is more important for Oracle because Oracle has
more alternatives to tree_sortkey (e.g., connect by) than Postgres
does.