The shelf below the shelf, and a verdict nobody was reading

VEILOS grades its own health in the open. The last release made sure every item on the top-level health report was accounted for — but that report has a second layer beneath it, and this release found that the accounting stopped at the first. Three health verdicts were being published there with nothing reading them, including one added only a release earlier: a statement about whether the system's storage measurements have been collected long enough to be worth reviewing at all. It was invisible to the existing check for three separate reasons at once, which is why the fix is a full accounting of that second layer rather than a wider search. Two of those verdicts also had a subtler flaw: a system that had observed nothing reported the same word as one that was halfway through observing, and an empty list of findings could mean either "nothing needs attention" or "nothing could be examined". Both now say which. Separately, the health grade itself had never been connected to anything: it was published, and no release check read it. The faults that this project can fix by editing its own code now hold up a release; the ones that merely report a fact about the world still do not, because refusing to ship over a condition would block the very release that fixes it. Money handling gained the same treatment — a card authorization left unresolved by an interrupted process is now examined by the system itself every hour and settled only from what the payment provider actually says, never from how long it has been waiting. Finally, two published promises were corrected: the privacy page said the site runs no browser code at all, which two pages quietly contradicted, and the security page pointed researchers at an address that no longer matched the machine-readable one. Both now say what the code does, and a test drives the real site to keep them honest.