Configure a dead-letter queue for messages that exceed the max receive count is correct because it moves messages that fail processing repeatedly into a separate queue for later inspection while letting normal traffic continue on the primary queue.

Using a dead-letter queue with an SQS redrive policy and a configured max receive count isolates so called poison messages and prevents them from repeatedly being retried by consumers. This approach preserves throughput for healthy messages and enables troubleshooting of payloads and error conditions in the separate queue without impacting normal processing.

Increase the queue's visibility timeout to 90 seconds can help reduce premature retries when processing takes longer than expected, but it does not provide a place to quarantine messages that will never succeed and it will not help you inspect problematic messages.

Use long polling to reduce empty receives lowers the number of empty responses and API calls, but it does not address messages that consistently fail after being received and it will not isolate those messages for later analysis.

Use a temporary response queue for each failed message is a pattern for request and response workflows and it is not a practical failure isolation mechanism for SQS consumers. Creating per-message temporary queues would be inefficient and it would not provide the centralized investigation point that a dead-letter queue offers.

Full AWS Practitioner Certification Question