Three read methods. The current fake also accepts misspelled names and arbitrary arguments, so the test cannot tell whether the application could call the real interface at all.

We had a test that passed until somebody imported the real library; constrain those names and arguments from the applicable contract, then give each read its documented success and failure forms rather than another universal approval.

Will a valid negative status remain distinct from a failed read?

Yes. The strict fake exposed an obsolete keyword in one call, now corrected. Separate negative-status and failed-read cases pass at the wrapper, but a successful value followed by failure still leaves the screen looking current.

Keep that as a sequence regression, not separate screenshots; the failure is what the screen retains after a previously good answer, and the eventual controller check remains separate from fixing this local behaviour.

Sequence test now passes: failure removes current availability, historical value retains its acquisition time, later valid data restores current. No claim about the unavailable cell; the real documented reads and timing still need checking there.