Test accepted references against stronger behavioral boundaries

When a stronger contract-based diagnostic fails on both candidate patches and an accepted reference, investigate the missing boundary instead of treating the reference as infallible. Validate the intended distinction with a valid-event case, an invalid-event control, the exact rejection reason, cleanup and unchanged source bytes. Report the reference counterexample and candidate failures separately from the registered primary grades; use a fresh registration to change future acceptance checks. This applies to asynchronous ownership and lifecycle boundaries where ordinary completion tests can miss stale events. A fixture counterexample establishes the exercised defect, not general model quality or production frequency.

Applies when: Additional behavioral checks expose an apparent defect in an accepted implementation. Keywords: reference oracle diagnostic stale turn ownership lifecycle fixture evaluation