Changed the software environment and our availability screen still looks reassuring, even when the request fails. The old SDK gave us a status value; this environment returns a container with more information. I'm comparing them on the maintenance bench. What should I inspect before I teach the screen another wrong interpretation?
Compare a saved successful response and a failed one with the documentation for that installed SDK. A plausible number inside a container is not necessarily the status you asked for.
Found our adapter taking the last numeric field regardless of the result indicator. On the failed example it's diagnostic information, not an availability value. So the screen is helpfully decorating an error with a status name.
Victor's question matters if both environments remain supported. I would select a documented adapter for each, with the same explicit available or unavailable result above them. Not guess the version from whatever fields happen to parse.
Victor and Leo, the old wrapper raises a documented read error; the new one returns a failed-result container. We've mapped both to unavailable and kept unexpected shapes as errors. Sam, startup has no value until a valid success. That check is now in the saved-response tests.
I would check that on the actual screen, Rosa. The parser can be right while a separate display cache leaves the old availability label looking current.
That's exactly where the next failure was. Parser said unavailable, screen kept its green last value. We've removed the current-state styling on failure; the optional previous reading now says last successful read with its time. Thanks Victor, the parser-only test would have missed it.
Both supported environments now pass startup failure, success then failure, recovery, and reopen checks with the saved examples. We also ran the normal documented read in each environment and checked the display transition on a controlled read failure. No diagnostic field becomes an availability state. Fix and handover accepted by our maintainer.