"Quentin de Metz" <quentin@de.me.tz> writes:
> It appears that changing the transaction_timeout when inside a transaction does not work as expected.
The effective value is whatever it was at transaction start, because
that's when the transaction timeout time is determined.
regards, tom lane