> The fact that the estimator knows that the LIMIT is pointless because
> there
> are less rows in the subselect than the LIMIT will return is not
> something we
> want to count on; sometimes the estimator has innaccurate information.
> The
> UNIQUE index makes this more certain, except that I'm not sure that the
> planner distinguishes between actual UNIQUE indexes and columns which are
> estimated unique (per the pg_stats). And I think you can see in your
> case
> that there's quite a difference between a column we're CERTAIN is unique,
> versus a column we THINK is unique.
I think a UNIQUE constraint can permit several 'different' NULL values...
better say "UNIQUE NOT NULL" ?