How can a pilot expose the limits of concurrent environment behaviour?

A retail evidence agent opens refund case REF-2187 and reads exception note EXC-0412 at version V-03. While the agent assembles its package, a store reviewer amends the note to version V-04 after finding that the recorded reason was incorrect.

The agent then submits a package referencing V-03. In the pilot's proposed behaviour, the environment changes the package to needs refresh and returns the revised version reference. The agent must inspect the amendment before deciding whether its conclusion still holds.

The consequence is a visible stale-evidence problem. The environment does not silently replace the cited text or accept a conclusion based on a version the agent never assessed.

This scenario helps a buyer set a meaningful scope limit. If concurrent actors are included, the pilot needs rules for when updates occur and how agents discover them. If the first release uses a fixed snapshot, the buyer should record that limit and assess concurrency separately.

The first step is to agree the expected submission response for this competing update. That decision makes the required state behaviour concrete before committing to a broader simulation of independent teams.