Steve King <steve(dot)king(at)ecmsys(dot)co(dot)uk> writes:
> The ctids are different however vaccum is run on this table and the record
> is updated.
It would be useful to look at xmin,xmax,cmin,cmax of these tuples too.
Actually, if you don't mind grabbing a copy of pg_filedump --- see
http://sources.redhat.com/rhdb/tools.html --- then a dump of page 7
of that relation would be good to look at. I am wondering about the
states of the infomask bits for these tuples...
regards, tom lane