Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> Remember one idea is for index scans to automatically update the expired
> flag in the index bitfields when they check the heap tuple.
There is no such flag, and I'm not planning to add one before I fix
statistics ;-)
There are a number of things that'd have to be thought about before
such a thing could happen. For one thing, trying to obtain write locks
instead of read locks during a btree scan would blow us out of the water
for supporting concurrent btree operations.
regards, tom lane