A single order event with an unexpected null crashed the consumer, went back to the queue, and was delivered again, forever. Behind it, thousands of orders waited. How we found it, why the queue had no redrive policy, and the max receive count, DLQ and redrive process we use now.
After we raised a Lambda timeout to handle large PDFs, customers started getting the same invoice email twice. The SQS queue's visibility timeout was still 60 seconds, so slow messages reappeared and a second invocation picked them up mid-processing.
SQS, SNS and EventBridge all move messages on AWS, and teams often pick whichever they used last. We compare them on what actually differs: pull versus push, fan-out, filtering, ordering, retries and replay, then show the combinations we use in production.
A new KEDA ScaledObject was meant to scale workers from zero on SQS queue depth. The queue reached 40,000 messages and the Deployment stayed at zero. The KEDA operator, not the worker, needed permission to read the queue. How we found it in the ScaledObject status and fixed it with EKS Pod Identity.