Konstantin Knizhnik <k.knizhnik@postgrespro.ru> writes:
> I wonder if it is considered to be expected behavior that concurrent
> execution of "create table if not exists" may return errors:
No, there's not a guarantee that IF NOT EXISTS will work in such
scenarios. As with everything else about that feature, it's a
quick-and-dirty, not very well designed effort.
regards, tom lane