When NMState is configured for updating DNS server, issue is observed on Egress Nodes, ordering of the IP Addresses changes, Node IPs are sorted by alphanumeric order.
Example: 
10.48.50.17 takes precedence of 10.48.50.2 (the actual host IP)

4.10.41

1. Install the cluster with OpenShift SDN
2. Configure projects with Egress IPs on specific nodes as per:
https://docs.openshift.com/container-platform/4.10/networking/openshift_sdn/assigning-egress-ips.html
3. Install NMState operator
4. Apply NNCP policy for updating DNS server:
~~~
apiVersion: nmstate.io/v1
kind: NodeNetworkConfigurationPolicy
metadata:
  name: dns-policy-mb 
spec:
  nodeSelector: 
    node-role.kubernetes.io/worker: ""
  maxUnavailable: 3 
  desiredState:
    dns-resolver: 
      config:
        search:
        - example.com
        server:
        - **.**.**.**
        - **.**.**.**
~~~
5. Try using egressIP, generate traffic from egressIP assigned namespace.

Primary IP address on Egress Nodes are sorted using alphanumeric order.

IP address ordering should not change on Egress Nodes.

At first no issue is observed, Node is Ready, egress IPs are assinged in the netnamespace and hostsubnets are visible, but connectivity is interrupted when traffic is generated from within egressIP assigned namespace.

Generate egress Traffic going through an egress IP after applying the NNCP
Compare the output of "ip a l" on a egress node before and after applying the NNCP