Claude Code reruns passing tests; a trace workflow quantifies the cost

On 2026-10-29, Golem.de published a hands-on test by Fabian Deitelhoff examining how Claude Code handles test runs during development. The key finding is that Claude Code can re-execute tests that have already passed, producing idle wait time and extra work that the developer did not ask for. Because those costs are easy to overlook in day-to-day work, the article proposes a workflow for observing them directly. Developers capture logs, metrics, and traces for every attempt the agent makes, then review that record to spot repeated failures, idle cycles, and additional work that do not move the task forward. The author positions the piece as a practical AI and testing experiment rather than a benchmark, so the numbers are tied to the author's own workflow rather than a standardized score. As of the publication date, the discussion remains at the level of a single practitioner's findings and the suggested logging workflow; no broader follow-up from other teams has been reported.