A staff misconfiguration prevented routes from being sent out to the upstreams, which has now been resolved.