Simon Riggs <simon@2ndQuadrant.com> writes:
> This would only happen if a VACUUM FULL had been run on the pre-9.0
> database and it had failed part way through.
If that were true, you might have an argument, but it isn't. VACUUM
FULL was never very careful about getting rid of all MOVED_xxx bits.
See the comments for update_hint_bits().
> We should remove the moved in/off flag bits and make it a part of the
> upgrade process to ensure the absence of those states.
That's not happening. The whole point of upgrade in place is to not do
anything as expensive as a full-database scan.
regards, tom lane
