There's a use case to replace CLUSTER, which requires downtime and needs to be rerun for new/updated data.