Oleg Bartunov <oleg(at)sai(dot)msu(dot)su> writes:
> test=# update tst set i = i+10 where a && '{3,4}';
> UPDATE 3267
> test=# set enable_indexscan=off;
> SET VARIABLE
> test=# update tst set i = i+10 where a && '{3,4}';
> UPDATE 4060
I get the same in current sources (in fact the number of rows updated
varies from try to try). Are you sure it's not a problem with the
gist index mechanism?
regards, tom lane