David Rericha <d(dot)rericha(at)healthcareoss(dot)com> writes:
> Thanks, Tom. That was it. After I ran analyze on the table, the error
> disappeared.
Yeah, I just duplicated the failure here. It only happens if the
comparison value falls into the histogram bucket adjoining the
out-of-range value, which is why you didn't see it happen for some
values of the comparison constant. Doesn't look too hard to fix.
regards, tom lane