Openshift 4.14+ embeds Multus 4, deployed by the Cluster Network Operator, using thick plugin mode.
That deploys a Multus daemonset, and the delegate CNI is launched within the Multus pods instead of running on the host. This brings several issues to  userspace CNI:

Multus daemonset uses a hostPath volume: / mount to /hostroot

This volume uses mountPropagation: HostToContainer option
CNI does bind mount of sockets from the Kubevirt virt-launcher pod to /var/run/vhost_sockets on the node
For this bound to be reflected to the node and then to the VPP pod, we need a mountPropagation: Bidirectional option

    Always (Limitation)

    1.
    2.
    3.
    

    Bidirectional mount propagation for the hostroot

Doug Smith raised this request through https://github.com/k8snetworkplumbingwg/multus-cni/pull/1290