A Testing Center run reports failures for the expected action even though the final answer sounds reasonable. What should the team do?

1. Inspect action-selection evidence and validate the expected outcome before changing the test
2. Delete every failed test without analysis
3. Publish immediately because an HTTP response was returned
4. Mark all textually fluent responses as passed