Recovering failed weekly support digests with dead-letter queues in Node.js

The article argues that choosing between Google Cloud Pub/Sub dead-letter retries and an AWS SQS dead-letter queue should depend on the recovery window, meaning how long a failed digest stays useful before support teams need it. Broker retry settings alone cannot tell whether a customer is still active or whether a digest was already delivered, so the application should carry that context. The recommended design makes retry rules explicit, using bounded attempts, idempotency keys, and a quarantine path for errors that won't clear. Digest state is tracked in the application database with leases so crashed workers don't cause duplicate sends. Recovery is scoped per customer, which lets operators replay a single digest without rerunning the whole cohort.