In a typical development environment, where metadata is frequently changing, having too many cluster nodes can reduce performance because of the need to synchronize them. In a production environment, when the metadata rarely changes, the impact of having many nodes is less likely to affect performance.