Tom Lane wrote:
> ...planner is actually going to choose based on the ultimate join cost,
> not on the subplan costs...
>
> In this explanation, the reason for the change in plans over time could
> be a change in the statistics for the other table. Is "facets" more
> dynamic than "point_features"?
In total rows changing it's more dynamic, but percentage-wise, it's
less dynamic (point_features probably turns round 50% of it's rows
in a day -- while facets turns over about 3% per day -- but facets
is 1000X larger).
Facets is a big table with rather odd distributions of values.
Many of the values in the indexed columns show up only
once, others show up hundreds-of-thousands of times. Perhaps
an analyze ran and just randomly sampled differently creating
different stats on that table?