> The sub-statements in WITH are executed concurrently That much I did get from the docs. Given each sub-statement is qualified by the RETURNING results of the previous one, that should at least guarantee the completeness of each query regardless of the order in which individual rows are affected.
Thanks, this has been helpful -- as I reread the page you pointed to, this jumped out at me: > All the statements are executed with the same *snapshot* (see Chapter 13 <https://www.postgresql.org/docs/current/mvcc.html>), so they cannot “see” one another's effects on the target tables. That I think specifically covers my question, and Tom Lane's note about end-of-statement was also helpful.
