Postgresql rank() with LIMIT 1 = 10x speed up?

index.sql

CREATE  INDEX  "index_patrons_on_amount_raised_in_cents" ON "patrons"  ("amount_raised_in_cents" DESC)

limit-subquery-1ms.sql

SELECT ranks.rank
FROM (
  SELECT id, rank() OVER (ORDER BY amount_raised_in_cents DESC) AS rank
  FROM patrons
) ranks
WHERE id = 1
LIMIT 1
        
                                                           QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=0.29..1429.42 rows=1 width=8)
   ->  Subquery Scan on ranks  (cost=0.29..1429.42 rows=1 width=8)
         Filter: (ranks.id = 1)
         ->  WindowAgg  (cost=0.29..1304.41 rows=10001 width=8)
               ->  Index Scan using index_patrons_on_amount_raised_in_cents on patrons  (cost=0.29..1154.39 rows=10001 width=8)
(5 rows)

(0.3ms)

no-limit-subquery-10ms.sql

SELECT ranks.rank
FROM (
  SELECT id, rank() OVER (ORDER BY amount_raised_in_cents DESC) AS rank
  FROM patrons
) ranks
WHERE id = 1
        
                                  QUERY PLAN
-------------------------------------------------------------------------------
 Subquery Scan on ranks  (cost=990.47..1290.50 rows=1 width=8)
   Filter: (ranks.id = 1)
   ->  WindowAgg  (cost=990.47..1165.49 rows=10001 width=8)
         ->  Sort  (cost=990.47..1015.47 rows=10001 width=8)
               Sort Key: patrons.amount_raised_in_cents
               ->  Seq Scan on patrons  (cost=0.00..326.01 rows=10001 width=8)
(6 rows)

(6.9ms)          

With 100,000 records postgresql uses the index with / without LIMIT 1. It's 10x faster with LIMIT 1.

I suspect that Postgresql doesn't know that the subquery returns unique IDs. Adding LIMIT 1 might make it stop scanning the subquery results on the first matching id.