HostedControlPlane Nodeport service is not opened in a dualstack deployment

After extensive debugging on HostedControlPlanes in dual stack mode, we have discovered that QE department has issues in dual stack environments. 

In Hypershift/HostedControlPlane, we have an HAProxy in the dataplane (worker nodes of the HostedCluster). This HAProxy is unable to redirect calls to the KubeApiServer in the ControlPlane, attempts to connect using both protocols, IPv6 initially and then IPv4. The issue is that the HostedCluster is exposing services in NodePort mode, and it seems that the masterNodes of the management cluster are not opening these NodePorts in IPv6, only in IPv4.
Even though the master node shows this trace with netstat:

tcp6 9 0 :::32272 :::* LISTEN 6086/ovnkube

It seems that it is only opening in IPv4, as it is not possible to connect to the API via IPv6 even locally. This only happens with dual stack; in the case of IPv4 and v6, it works correctly in single-stack mode.

4.14.X
4.15.X

100%

1. Deploy an Openshift management cluster in dual stack mode
2. Deploy MCE 2.4
3. Deploy a HostedCluster in dual stack mode 

- Many pods stuck in ContainerCreating state
- The HostedCluster cannot be deployed, many COs blocked and clusterversion also stuck

HostedCluster deployment done

To reproduce the issue you could contact @jparrill or @Liangquan Li in slack, this will make things easier for the environment creation.