Gregory Stark <stark@enterprisedb.com> writes:
> I had another thought. Perhaps in use_assert_checking mode we should have it
> start from a random position every time. Or perhaps just a random position
> from amongst the first n pages.
Interesting idea, but I fear it'd make a lot of the regression tests
fail. Might be worth trying to see how bad things are there, though.
regards, tom lane
