All comments spot on! I've been advocating for better batching support in Kafka Streams for at least 6 years. You can do it with the current version of Kafka but it is oh so painful. Confluent is positioning Flink as the answer to batching needs but Flink is not a good fit for a microservice architecture.