What happens when a system is temporarily unreachable?

Outages of individual systems are normal in operations, whether from a maintenance window, an update or a network glitch. Our middleware architecture is built for exactly this scenario: messages are persistently stored in message queues before they are delivered. If a target system goes offline, no orders or stock movements are lost; they wait in the queue.

While a system is unreachable, automatic retries run with exponentially growing wait times (backoff). A circuit breaker prevents the struggling system from being flooded with requests. After recovery, the middleware delivers all pending messages in the correct order. Anything that still cannot be delivered after the defined attempts lands in a dead-letter queue for manual review, instead of disappearing silently.