Message Queues

Design and diagnose asynchronous delivery without inventing guarantees the broker or application does not provide. Make the delivery contract explicit: who owns a message, when an effect is safe to repeat, how poison messages stop blocking progress, and how recovery avoids replaying more than intended.

When to use

• Designing Kafka, RabbitMQ, or compatible producer and consumer delivery behavior
• Investigating consumer lag, growing queues, duplicate effects, redeliveries, or stuck messages
• Choosing acknowledgement, retry, dead-letter, ordering, backpressure, retention, or replay rules
• Reviewing an event-driven workflow for idempotency, observability, and failure recovery

When NOT to use