Kafka Share Groups - Pathological fetch waits with record_limit

This blog post discusses performance issues in Kafka Share Groups when using the `share.acquire.mode=record_limit` setting with fewer consumers than partitions. It examines how partition skew can lead to inefficient performance and highlights specific tests that consistently fail in this configuration. The post aims to explain the struggles observed during testing and provides insights into optimizing throughput while maintaining target latencies.