What were you trying to do that didn't work?

At our hypershift hosted cluster we have some states at MCO to manually enable the arp proxy default gw 

https://github.com/openshift/machine-config-operator/blob/master/templates/common/kubevirt/files/002-nmstate-arp-proxy-ipv6-gw.yaml

# cat /etc/nmstate/002-nmstate-arp-proxy-ipv6-gw.applied 
capture:
  ethernet-nics: interfaces.type=="ethernet"
desiredState:
  routes:
    config:
    - destination: ::/0
      next-hop-interface: "{{ capture.ethernet-nics.interfaces.0.name }}"
      next-hop-address: fe80::1

After applying it we see the following log from NetworkManager

audit: op="connection-update" uuid="ccb86a95-fda5-3b2b-9349-f745705ed8b3" name="Wired connection 1" args="ipv4.dhcp-client-id,ipv4.dhcp-timeout,ipv4.addresses,ipv6.addresses,ipv6.routes" pid=1942 uid=0 result="success"

And the ipv4 address that was already received from the DHCP server get configured as if it was an static one and appear at ipv4.address

This breaks us since the interface appear with a pair of addresse and the same subnet and that breaks kubelet.

Please provide the package NVR for which bug is seen:

How reproducible: Always

Steps to reproduce

Simpler reproducer:
1. Setup a DHCP environment.
2. Enable DHCPv4 and DHCPv6 on DHCP interface.
3. Wait DHCP process done and note down auto IP address.
4. Apply this state

---
routes:
  config:
    - destination: ::/0
      next-hop-interface: dhcpcli
      next-hop-address: fe80::1

5. Check whether previous auto IP address is dynamic(not forever `valid_lft`).
6. Check `nmcli c show dhcpcli` to make sure no auto IP address been stored as static ip into `ipv4.addresses`.

Expected results

   the ipv4.addresses attribute should not be set since the addreses are from DHCP server and not static

Actual results

    the the NetworkManager ipv4.addresses attribute is set.