A security team built an automated incident-response pipeline: EventBridge routes GuardDuty findings to an SQS queue, and a Lambda function processes each message to isolate compromised instances. During a large-scale incident, GuardDuty generated thousands of findings simultaneously. The team discovered that one malformed finding caused the Lambda function to throw an exception repeatedly, and the same message was reprocessed continuously while valid isolation actions were delayed. What is the BEST way to prevent a single failing message from blocking processing of other findings?