Create a hostedcluster, the version is 4.12.16, the nodepool version is 4.12.16, then upgrade nodepool version to 4.13.0-rc.8, upgrade should failed, but the existing machine should not be deleted, and the machine with 4.13 should not be provisioned.

quay.io/openshift-release-dev/ocp-release:4.12.16-x86_64
quay.io/openshift-release-dev/ocp-release:4.13.0-rc.8-x86_64

1.Always

1.Create a hostedcluster with 4.12.16. Waiting install successfully.
2.Upgrade nodepool version to 4.13.0-rc.8
3.Check machines resources. 

[hmx@fedora hypershift]$ oc get machines -A
NAMESPACE                 NAME                                         CLUSTER                NODENAME                                    PROVIDERID                              PHASE      AGE     VERSION
clusters-mihuangtest517   mihuangtest517-us-east-2a-5bc9f99c98-d9qr8   mihuangtest517-lhq94   ip-10-0-141-78.us-east-2.compute.internal   aws:///us-east-2a/i-05d98127ce339440e   Running    4m41s   4.13.0-rc.8
clusters-mihuangtest517   mihuangtest517-us-east-2a-6766499477-4cmrp   mihuangtest517-lhq94   ip-10-0-135-50.us-east-2.compute.internal   aws:///us-east-2a/i-06e3c2d39e7ddd62b   Deleting   104m    4.12.16
[hmx@fedora hypershift]$ oc get machines mihuangtest517-us-east-2a-5bc9f99c98-d9qr8 -n clusters-mihuangtest517 -o yaml
......
  conditions:
  - lastTransitionTime: "2023-05-17T05:19:49Z"
    status: "True"
    type: Ready
  - lastTransitionTime: "2023-05-17T05:19:31Z"
    status: "True"
    type: BootstrapReady
  - lastTransitionTime: "2023-05-17T05:19:49Z"
    status: "True"
    type: InfrastructureReady
  - lastTransitionTime: "2023-05-17T05:23:03Z"
    status: "True"
    type: NodeHealthy
  infrastructureReady: true
  lastUpdated: "2023-05-17T05:22:24Z"
  nodeInfo:
    architecture: amd64
    bootID: a9114a01-51a1-456e-b06e-e302dc58fefe
    containerRuntimeVersion: cri-o://1.26.3-3.rhaos4.13.git641290e.el9
    kernelVersion: 5.14.0-284.13.1.el9_2.x86_64
    kubeProxyVersion: v1.26.3+b404935
    kubeletVersion: v1.26.3+b404935
    machineID: ec2410ca94ffd8daefc6c1c195075a39
    operatingSystem: linux
    osImage: Red Hat Enterprise Linux CoreOS 413.92.202305041429-0 (Plow)
    systemUUID: ec2410ca-94ff-d8da-efc6-c1c195075a39
  nodeRef:
    apiVersion: v1
    kind: Node
    name: ip-10-0-141-78.us-east-2.compute.internal
    uid: 61b133ba-2d8c-433b-aea0-e4252ccc9fa7
  observedGeneration: 2
  phase: Running

If an invalid machine version was set, the existing machine should not be deleted since we already set the condition to notify the users and cancel/block the upgrade operation, and  the machine with 4.13 should not be provisioned since our condition is invalid for this version.