I'm working on a performance test using Psycopg2 + Postgres. My workload is to do a spatial join to use one data point joining two reference tables by repeatedly issuing a query string with different coordinates. The average query execution time that I got is 0.5 ms, which is surprisingly fast. I started a pgsql terminal and analyzed the same query multiple times. The overall execution time is around 1.5ms with ~1ms in planning and 0.5ms in execution. That makes me is that because psycopg2 implicitly cached the plan. Is there any implicit optimization there? Thanks in advance!