Would love to hear any technical history you can share, what did this replace, and what shortcomings led to it being built? (Seems quite similar to Kafka but perhaps more emphasis on low latency rather than raw throughput?)