Matt <matt@ymogen.net> writes:
> Anyway, I was about to respond to Tom's comment that it looked like a
> snapshotting bug, but can we take it from yr subsequent post Tom that
> you no longer think so?
Right, assuming that what Murthy is seeing matches the behavior I
described (ie, no more than a few zeroed pages, all near the ends of
tables that were undergoing extension).
regards, tom lane