Multiplets, Models, and the Search for Meaning: Improving Per-Test Fault Diagnosis