When the server is booting, repro may be started while PostgreSQL is still initializing.
It fails with an error like this:
PostgreSqlDb.cxx:173 | PostgreSQL connect failed: FATAL: the database system is starting up
It needs to be more tolerant of connection (and reconnection) issues and operate sensibly during temporary PostgreSQL unavailability.
In case of hard errors (e.g. an invalid PostgreSQL hostname) it is probably OK to just stop the repro process completely as it does now.