On Thu, 23 Sept 2021 at 13:28, David G. Johnston <david.g.johns...@gmail.com> wrote: > Agreed, though if the query author needs to do that here we’ve violated the > spirit of the declarative SQL language. At first blush nothing about the > original query seems like it should be preventing parallelism. Each worker > builds its own distinct array then the final concatenation is made distinct.
We don't parallelise DISTINCT / ORDER BY aggregates. David
