The clusterctl CLI will create all the four needed components cluster-api (CAPI), cluster-api-bootstrap-provider-kubeadm (CAPBK), cluster-api-control-plane-kubeadm (KCP) and cluster-api-provider-hetzner (CAPH). It uses the respective *-components.yaml from the releases. However, these are not highly available. By scaling the components we can at least reduce the probability of failure. For whom this is not enough could add anti-affinity rules and PDBs.
Scale up the deployments
kubectl -n capi-system scale deployment capi-controller-manager --replicas=2
kubectl -n capi-kubeadm-bootstrap-system scale deployment capi-kubeadm-bootstrap-controller-manager --replicas=2
kubectl -n capi-kubeadm-control-plane-system scale deployment capi-kubeadm-control-plane-controller-manager --replicas=2
kubectl -n caph-system scale deployment caph-controller-manager --replicas=2