The event
What the room explored
The sessions shared a common thread: abstractions that appear simple at the API boundary become difficult when ordering, failure, recovery, and throughput meet. Kafka timestamps influence correctness, consumer ownership shapes concurrency, and a filesystem connector can determine the cost of every Flink checkpoint.
Rather than treating those as isolated configuration problems, the speakers traced them back to system design. The talks showed how newer Kafka and Flink capabilities can remove old compromises while making the operational contract more explicit.