I thought it would be helpful to quickly point out in this thread that I amnot imagining things, and that Tom Lane has confirmed inpgsql-performance that this is a genuine issue:
https://www.postgresql.org/message-id/17679.1487683929%40sss.pgh.pa.us