How can an evaluation expose routing behaviour that keeps work moving without resolving ownership?

An insurer’s intake agent routes REQ-0631 to claims administration because the message mentions a repair invoice. Claims administration returns it with REJECT-0148, explaining that the invoice concerns a supplier account rather than a policyholder claim. The request returns to intake with that rejection linked to its history.

The agent then sends REQ-0631 back to claims administration without obtaining new evidence or changing its rationale. The queue rejects it again. Each transfer is allowed by the tool, but the case has entered an unproductive cycle and no team has accepted responsibility for the supplier enquiry.

The evaluator can inspect whether a repeated destination was supported by changed evidence. It can distinguish a justified retry after correcting a missing reference from repetition that ignores a recorded rejection. A separate check can identify whether the agent used the supplier-account clue to investigate an appropriate destination.

MirrorFirm could generate rejection histories and tasks that begin partway through this cycle. The scenario teaches builders to assess progress across the trajectory: counting successful tool calls or checking only the latest queue would miss the operational failure.