[4.10.z] MCP fails to update after SR-IOV drains the node

After applying manifests for SR-IOV configuration MCP fails to update

4.10.60

100%

1. Deploy baremetal OCP cluster
2. Create MCPs - 1st for regular worker nodes and 2nd for SR-IOV workloads
3. Apply manifests:
---
apiVersion: sriovnetwork.openshift.io/v1
kind: SriovOperatorConfig
metadata:
  name: default
  namespace: openshift-sriov-network-operator
spec:
  configDaemonNodeSelector:
    node-role.kubernetes.io/sriov: ""
  enableInjector: true
  enableOperatorWebhook: true 

---
apiVersion: sriovnetwork.openshift.io/v1
kind: SriovNetworkNodePolicy
metadata:
  name: "sriovleftdpdkmellanox"
  namespace: openshift-sriov-network-operator
spec:
  resourceName: "sriovleftdpdkmellanox"
  nodeSelector:
    node-role.kubernetes.io/sriov: ""
  mtu: 9000
  numVfs: 4
  nicSelector:
    # consider switching to PCI paths
    pfNames: ['ens4f0', 'ens5f0']
  deviceType: netdevice
  isRdma: True

Node stuck in SchedulingDisabled and MCP is marked as `paused`

SR-IOV is successfully configured and nodes are rebooted

Notes from Sebastian:
---------------------
the problem is now we need to drain also after a reboot when the number of devices is 0
but we are not able to lock the drain as in the reboot time another node already took the lock
so we are in a dead lock