Can generated escalation data include awkward cases, conflicts and unavailable approvers?

**Can a release include conflicting policies?**

Yes, a buyer can ask for it. For example, a policy document may state a limit that differs from the structured authority table, or two policy versions may both appear current. The task pack would label which source is authoritative, or whether the expected behaviour is to flag the conflict.

**What about an approver who is not available?**

A scenario could mark the named senior role as on leave, with a delegate listed in a separate record. The agent must find the delegate rather than stall or choose the wrong person.

**Can requests be split to stay under a limit?**

A dataset could include related requests sharing a party, loss event or order, where each is under the limit and the total is not. This lets a grader test aggregation.

**Can evidence be incomplete?**

Yes. Cases can omit a document the senior reviewer would need, with the expected behaviour being to gather it first or state the gap in the handover note.

**Are these labelled?**

A useful pack attaches an expected outcome and the reason to each case, so success does not depend on a grader guessing intent.

**What are the limits?**

The cases are generated, so coverage of rare real-world situations depends on what you scope. Inspect grounding and disclosure risk, and use a pilot to confirm the awkward cases that matter to you.