Making retries safe: idempotency keys, backoff, and dead-letter queues

Networks fail in the middle of requests, so real systems retry, and retries are where duplicate charges and lost messages come from. Learn the difference between at-least-once, at-most-once, and 'exactly-once' delivery, and why the last one can't exist over a lossy network. You'll also cover idempotency keys and server-side dedup, exponential backoff with jitter, which requests are safe to retry, the Two Generals' Problem, and poison messages, plus how to retry a payment without charging the customer twice.