If a table has no indexes, we will always decide that any same-page
update operation is a HOT update, since obviously it isn't modifying
any indexed columns. But is there any benefit to doing so? I don't
see one offhand, and it has a downside: we're very likely to
encounter broken HOT chains if an index is created later. That leads
to the sort of unexpected behavior exhibited here:
http://archives.postgresql.org/pgsql-performance/2010-11/msg00216.php
I'm thinking maybe HeapSatisfiesHOTUpdate should be changed so that it
always returns false if the relation has no indexes, which could be
checked cheaply via relation->rd_rel->relhasindex.
regards, tom lane
