There's some work needed in Cluster API upstream which is critical for supporting desired NodePool behaviour and value.

1 - Bubble up conditions from infraMachines resources.

This is critical to signal failure scenarios to consumers.

kubernetes-sigs/cluster-api#6218
kubernetes-sigs/cluster-api#6025

We are workarounding this with https://github.com/openshift/hypershift/pull/1907.

This is suboptimal implementation for multiple reasons, e.g. it does not scale well for huge NodePools, it misses meaningful messages from the infraMachine, e.g

The infraMachine shows

```

 - lastTransitionTime: "2022-11-30T13:05:21Z"
    message: "failed to create AWSMachine instance: failed to run instance: InvalidParameterValue:
      Value (20756lq9aha19chgodjd8krdtr49q8o6-worker-profile) for parameter iamInstanceProfile.name
      is invalid. Invalid IAM Instance Profile name\n\tstatus code: 400, request id:
      cda1113b-2f2d-4c33-a46e-961401ec03c7"
    reason: InstanceProvisionFailed

```

But the workaround can only consume from Machines which show

```

  - lastTransitionTime: "2022-11-30T12:46:36Z"
    message: 0 of 2 completed
    reason: InstanceProvisionFailed
    severity: Error
    status: "False"
    type: InfrastructureReady

```

2 - Propagate labels

This is critical UX tag pool of Nodes and keep it after replace upgrades. A common use case is to tag a pool as "infra" which directly affects to cost.

https://github.com/kubernetes-sigs/cluster-api/pull/7173