It accepts invented method names. We have saved replies, but no version recorded with them yet.

Keep those captures, but label the version gap. They can help investigate the parser without becoming the authority for the intended interface. Compare against the actual installed API contract before calling the substitute faithful.

When I've handed over test results, people read passed and stop there; who will see exactly which response cases you exercised and which are still missing?

Grace, the test names should say that plainly, not hide the cases in the mock's internals; a report headed robot tested would be nonsense here

Application maintainer gets the case list. Scope says parser and display only. Version identified from the original capture environment now.

Include failure after a successful response. A parser can reject the failure correctly while the screen keeps showing the earlier value as current.

Also distinguish a valid negative fixture state from a failed request. Both may contain something false, but one is an observation and the other leaves you without a current observation.

That distinction found a bug: request failure displays fixture absent. It should show no current reading, not a physical judgement.

Keep that example as a regression case. Otherwise someone can simplify the two states back into one boolean later and admire the shorter code

Fixed in the offline version. Valid absent and failed read display differently. Thanks Rachel. Malformed responses and stale-value presentation still need tests.

Leave those gaps on the handover list. The corrected example is useful evidence about the application; comparison with the real controller remains another step.