This is not a replacement for node failover, but instead provides a stop-gap for gracefully handling situations where a particular document is inaccessible but available through one of its replicas. It is worth noting that this feature does not guarantee consistency as any pending writes on the primary node may not have been replicated yet.