We can migrate a cluster created with a single arch payload to a multi payload using the `oc adm upgrade --to-multi-arch` command. While the migration is happening, if we simultaneously provision a machineset (with the appropriate arch specific bootimage) of a different architecture (differing from the control plane arch), the machine stays in provisioned state for ever and no node is created.

This is due to the fact that the machine does get created and boots up, but when MCO pivots to the machine-os-content, it pivots to the single arch machine-os because the upgrade is not complete yet.

While this case is indeed rare, it would be great if somehow this error could be propagated out of the machine and it would transition to a failed state 

4.13.0-ec.4

always

1.Create an AWS amd64 cluster with a single arch 4.13.0-ec.4 payload 
2. Execute the migration command `oc adm upgrade --to-multi-arch`
3. Provision a machineset with the arm64 bootimage and instance type
4. monitor with `oc get machines -n openshift-machine-api`

machine stays in provisioned state

machine transitions to failed state

logs from machine-api-controller:
I0314 23:48:15.684256       1 reconciler.go:407] psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr: ProviderID already set in the machine Spec with value:aws:///us-east-1c/i-0c0d4fb0fe65f651d
I0314 23:48:15.684315       1 reconciler.go:267] Updated machine psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr
I0314 23:48:15.684323       1 machine_scope.go:167] psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr: Updating status
I0314 23:48:15.761009       1 machine_scope.go:193] psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr: finished calculating AWS status
I0314 23:48:15.761027       1 machine_scope.go:90] psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr: patching machine
I0314 23:48:15.778043       1 controller.go:341] psundara-mycluster01-8mrjn-worker-us-east-1c-m5nlr: has no node yet, requeuing