Bugs
Can't inline set returning functions with declared out parameters
http://archives.postgresql.org/message-id/11061.1320352516@sss.pgh.pa.us
Unspecified join selectivity bug fixed in 9.0.5
http://archives.postgresql.org/message-id/12845.1319640329@sss.pgh.pa.us
removal of n_distinct clamping heuristic backfires in some cases
http://archives.postgresql.org/message-id/24198.1325284143@sss.pgh.pa.us
EC matching busted in some MergeAppend-related cases
Planner is too reluctant to do clauseless joins - fixed in 9.2
http://archives.postgresql.org/message-id/4F875293.1000100@nuxeo.com
One version of RHEL kernel had cache management bug
http://archives.postgresql.org/160356272.159746.1332878777645.JavaMail.root@mail-1.01.com
Allow child-relation entries to be made in ec_has_const EquivalenceClasses
Index size fudge factor
http://www.postgresql.org/message-id/51656E03.7000701@ca.afilias.info
http://www.postgresql.org/message-id/50EEC2D5.3050602@wp.pl [maybe]
http://www.postgresql.org/message-id/20121212234732.80090@gmx.com
Planner overestimates partial index rows to be visited - fixed in 9.2
http://www.postgresql.org/message-id/28778.1355972790@sss.pgh.pa.us
Planner overestimates pages to be visited in a bloated index
http://www.postgresql.org/message-id/50D8A0D7.2050005@richardneill.org [maybe]
Plan cache changes in 9.2 cause more object copying