In the production CI Mihawk with RHEL 8.8, worker nodes fail to come up in some of the job runs. Sometimes both worker nodes are missing, and sometimes only one.

Checked the machine api logs and this is what is found
I1026 09:43:44.109891       1 controller.go:278] libvirt-ppc64le-0-0-7-w7bg7-worker-0-bzd6p: reconciling machine triggers idempotent update
I1026 09:43:44.109901       1 actuator.go:189] Updating machine libvirt-ppc64le-0-0-7-w7bg7-worker-0-bzd6p
I1026 09:43:44.113090       1 client.go:142] Created libvirt connection: 0xc000a32c40
I1026 09:43:44.113370       1 client.go:302] Lookup domain by name: "libvirt-ppc64le-0-0-7-w7bg7-worker-0-bzd6p"
I1026 09:43:44.113719       1 actuator.go:364] Updating status for libvirt-ppc64le-0-0-7-w7bg7-worker-0-bzd6p
I1026 09:43:44.114695       1 actuator.go:506] The domain does not have any network interfaces
E1026 09:43:44.114704       1 actuator.go:385] Unable to get node addresses: requeue in: 1s

Looks like the worker nodes are not getting IP assigned

So far observed only on OCP 4.12 and below