First, this includes the pre-computation of the mirror responsible for a given key. For more information, you can see our blog posts Queryable Kafka Topics with Kafka Streams and Scaling Requests to Queryable Apache Kafka Topics with nginx.
We also have further minor inefficiencies in mind that we want to remove.