Flink platforms, state internals, watermarks, exactly-once — the compute layer above the broker. The Kafka alumni ladder.
Deliberately compute-layer: where the Kafka cohort taught the broker, this one teaches the processing — LinkedIn's 4-trillion-event pipelines and −94% unification, Alibaba's 4B records/sec Double-11 spine, and the state/time semantics that decide whether streaming results are actually correct.
Three exits from Lambda — collapse to streaming with a reprocessing ladder, one codebase over two runtimes, one engine running batch as bounded streams — each an insurance policy with premiums.
The feature ships twice. Every bug fix ships twice. The compliance change last quarter shipped twice, and the two implementations disagreed for three weeks before anyone noticed — the batch layer said one number, the speed layer another, and the dashboard quietly preferred whichever wrote last. Your team maintains one product surface backed by two codebases in two frameworks with two on-call rotations, and the architecture document calls this *a design*, citing a decade-old blog post. Everyone hates it. Nobody can say precisely what would break if the batch layer just… stopped. Could you?
The full week 4 brief is part of LeetData Pro.
