Pick the right error-handling strategy

Your team runs a Lambda function that is invoked asynchronously by Amazon S3 whenever a new object is uploaded. Occasionally the function fails on a specific event, and after Lambda exhausts its automatic retries the event is silently dropped. The team wants every event that fails all retries to be captured somewhere durable so it can be inspected and reprocessed later, with the least amount of custom code.

Which error-handling strategy best captures these failed asynchronous events?