A stalled shutdown now names what actually stalled

When something inside the platform takes too long to shut down, it writes a report so an operator can see what is holding it up. That report used to end with the same sentence every time: the hold-up is in something below this — go and look there. On one restart it said that 47 times about parts that had nothing below them at all, so every reader was sent to the wrong place.

Three of the numbers it printed were not measurements. One was a constant that read the same in every possible state. One was printed under a name that meant the opposite of the value. One was the report quoting its own starting point back as though it were progress.

Those are now real readings, and there is a new verdict for the case that had none — the one where the work is queued and simply never picked up. When none of the known causes fits, the report now says plainly that it does not know, and prints what it did observe, instead of picking the nearest-looking answer.

This makes the next occurrence diagnosable. It does not by itself stop the stall.

Reconnecting…
The server was updated. Reloading the page to pick up the latest version.