Tom Lane wrote:
> Your test case simply shows that it takes a finite amount of time for
> the previous backend to exit; depending on scheduler priorities it's
> possible for the next instance to connect and try to CREATE DATABASE
> before the previous one is gone. My suggestion would be to connect
> to some other database besides the one being copied.
>
>
>
Didn't we invent the "postgres" database for just this reason?
cheers
andrew