We've paused using the diagnostic display for decisions. After the SDK change, a failed read can still produce a green status because the wrapper accepts the returned container as a value. Need the interpretation fixed, not another reassuring screen colour.
Capture the successful and failed response structures for the installed SDK version. Have the wrapper check the documented success result before interpreting the payload. Add malformed responses to the offline cases as well.
Developer found a truth test on the whole response. Nonempty failure responses passed it. They've added explicit interpretation using the matching documentation and saved examples.
Test a successful read followed by a failure. The old value may still be useful as last known, but its original update time must not become the failed request time.
That test found the timestamp moving on failure. Fixed now: unavailable for current status, retained value clearly dated. Initial failure has no made-up zero. Operator wording review is next.
Did the operator review distinguish unavailable from last known without prompting? That is the remaining check I would want before relying on the display again.