PG Bug reporting form <noreply@postgresql.org> writes:
> Below is an example of queries. It shows that the query planner chooses
> sequential aggregation instead of parallel, although limit should not have
> affected this.
On what grounds do you assert that LIMIT can be parallelized?
AFAICS it has to be done by a single process to ensure that the
limit is enforced accurately. So LIMIT below an aggregation
removes the opportunity to parallelize the aggregation.
regards, tom lane