Have the author use the documented success and payload fields for the actual query in each supported package, because recognising a familiar-looking number is not enough to know the read succeeded

A failed read needs its own outcome. It must not become a machine mode, and it should not look like an inspection rejection either

Are both package versions meant to remain in use?

Yes, on separate maintained workstations for now; the author is defining an explicit configuration for each, with unsupported configurations refused rather than decoded by shape.

What happens to the last successful mode when a read fails? A covering operator may find it useful, but only if the display makes its age obvious.

Keep current observation unavailable and label any retained mode as previous. A fresh timestamp beside old data is especially misleading

Offline success-then-failure checks now show unavailable as current, with the previous successful observation and its original time separately labelled; malformed responses follow the same unavailable route.

And an unknown package?

Refused before interpretation in the test; there is no attempt to find a plausible mode inside its response.

Then retain those consecutive-response cases with both supported configurations, so the older workstation does not become the untested one simply because the replacement caused this complaint

Has cover read that previous-value display without its author explaining it? The separation can be technically right and still look like two competing current modes.

That walkthrough and the installed comparisons are still due; the offline parser checks are complete, but I have not issued this as a finished maintenance screen.