qwen-code adds review lenses after four passes missed a bug

qwen-code merged two new review checks on August 22, built from a bug that four rounds of automated review missed. The gap matters for any team treating repeated automated approval as a safe signal to merge. Passing review four times didn't mean the code worked.

PR #9708 targets one failure shape, whether a value exists at the moment its consumer needs it. The temporal-reachability lens asks that question directly, formatting findings as produced at X, needed at Y, Y precedes X. The incident-replay lens takes a different angle. It reads the failure story in a pull request description and checks the fix against it as a spec.

Both checks trace to PR #9655. Four rounds of automated review passed that code before a human found the one blocking defect eleven minutes after the last LGTM.

The pull request doesn't say how often the temporal-reachability check fires on qwen-code's codebase since merging. It also doesn't say whether the check throws false positives on correct code that's just ordered unusually.