Maxim Boguk <maxim.boguk@gmail.com> writes:
> I know GIST on intarray[] do not have that problem.
> Very likely the problem is limited to intarray[] GIN indexes only
> (but I going to test some other not-well known GIN indexes tomorrow).
> Broken FTS indexes on Hot Standby should be known years before.
You might think that, but you'd be wrong :-(. ginRedoUpdateMetapage
is failing to restore the contents of the pending-list correctly,
which means this is broken for all types of GIN indexes. Will fix.
regards, tom lane