We burned 70% of our supplier API quota confirming that nothing changed.

That was the moment we knew polling-based inventory sync had to go.

Just published a breakdown of why we moved away from REST for travel inventory sync — covering the architecture switch, the Kafka event schema we landed on, consumer design gotchas (idempotency, DLQ, order tolerance), and the results: booking failure rate dropped from 2.3% to 0.4%.

Also included a section on when REST is still the right call — because event-driven is a trade-off, not a default.

👉 [Why we stopped using REST for travel inventory sync, and what we use instead]

Would love to hear how others are handling this at scale — polling, push, hybrid?