If the root and/or DMZ relays have fairly static IPs, then it is not a bad idea to define some failover relays by IP address in addition to hostname.
This not only matters in this particular case, but also if there is a failure with DNS somewhere.
This also provides an interesting option for selecting local relays. You could define the first relay to be selected to be 192.168.1.200 which could be a different physical relay for every site using NAT with this subnet.
Given the odd configuration of this customer, it might make sense to have a relay in their network just to make it a bit easier for the clients to get to a relay.