Waiting for processing to finish before judging failures

This anonymized case study is currently maintained in Japanese. It explains how completion and failure counts were separated so that an in-progress run would not be judged as finished.