When services within the fat container try to communicate with each
other, like backend to RTS etc., if they use the loopback address of
IPv4, `127.0.0.1`, it works. But if they use the loopback address of
IPv6, `::1`, it fails because the NGINX inside the fat container isn't
set to bind to IPv6.
This PR fixes this.
In EE, we attempt to make connections to Keycloak without setting the
hostname on the `WebClient`. This picks up the hostname of `::1` on
systems with IPv6 as default, and so the communication between backend
and Keycloak fails.
This is affecting users on ECS Fargate, for example.
b3dbd85dbe