On controller replacement with predictable IPs, new controller ends up without IP addresses in Red Hat OpenStack Platform

Issue

On controller replacement with predictable IPs, new controller ends up without IP addresses for statically configured IPs in Red Hat OpenStack Platform.

This happens after following https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/8/html/director_installation_and_usage/sect-replacing_controller_nodes in an environment that uses ips-from-pool-all.yaml.

On the new stack update, the network deployment is timing out and the new controller's PXE / control plane interface is not reachable.
In order to investigate further, the system was booted into single user mode to change its password: https://access.redhat.com/solutions/918283
- additionally, the redirection from tty0 to console was removed from the kernel command line, otherwise no output would be shown once the shell was reached.

In /etc/os-net-config/config.json, instead of showing valid ip_netmask combinations, it only shows a netmask, e.g. /24:

{"type": "vlan", "addresses": [{"ip_netmask": "/24"}], "vlan_id": 902}]}

instead of

{"type": "vlan", "addresses": [{"ip_netmask": "172.16.0.6/24"}], "vlan_id": 902}]}