How One Engineer Solved Identity Loss Across Async Kafka Event Boundaries

In event-driven Spring Boot architectures, user identity validated at the API gateway can silently disappear once a request crosses into asynchronous processing. A developer writing for DEV Community explains that using the transactional outbox pattern — while correct for reliable event delivery — means the scheduled poller thread that publishes to Kafka has no access to the original caller's security context. Common workarounds like MDC or ThreadLocal storage fail because the publisher runs on a separate thread, long after the HTTP request has completed. This becomes especially problematic in dead-letter queue scenarios, where knowing who triggered a failed event is critical for debugging and forensics. The author's solution involves making the acting user's identity an explicit, durable part of the outbox record itself, so it travels with the event and survives retries and replays.