VF GUID Configuration in Nvidia Infiniband not being detected by SRIOV

After configuring this use case (Nvidia Infiniband on a Red Hat OpenShift connected or air-gapped cluster: https://www.redhat.com/en/blog/nvidia-infiniband-red-hat-openshift-connected-or-air-gapped-cluster), so activating MOFED using NVIDIA. There is an issue with vf GUID, when the mofed driver loads it gives all the vf 0, the SRIOV operator doesn't get the correct GUID configuration. 

The SRIOV operator should catch the network cards, configure VFs and assign GUID on the cards. There is possible race condition, where the MOFED pods are coming up, only if restart of the pods of the SRIOV operator (sriov-network-config-daemon and sriov-device-plugin) then the pods are working well.
It looks like the SRIOV pods are checking the PCI cards only when coming up, and not during runtime.
Currently the system is working after restarting the pods.

    4.12

We can not collect SRIOV must-gather (high security infra) nor we do have labs for checking further to reproduce from our end in support.    

As temporary workaround the team needs to restart the SRIOV pods manually during mofed pod loading and the vf gets the correct GUID.    

    Not needed to restart pods.

    OCP 4.12 , DGX of 8xA100-80G each server