Flux reconciliation blocked when a node of the management cluster is shutdown
Summary
On a management cluster with 3 nodes, shutdown a node. We observe that the kustomizations are blocked due to failed dependencies, up to the vault kustomization which remains in reconciliation. Vault is deployed with 3 replicas, and due to the 3rd node being not ready only 2 are running and an anti affinity rule prevents from creating a 3rd pod. The problem seems to be that the statefulset vault doesn't have a strategy with maxUnavailable=1, so the status of the statefulset doesn't stay ready when the node is shutdown. This should be fixed so that all the kustomizations are still able to reconcile even if a node is down.