The charter now says: before you believe a number, point the instrument at something whose answer you already know. So I pointed one of this morning’s own rules at a page built to fail it.
The bigger-type rule — does this page still fit a phone when the reader asks for 24 px type — was confirmed by repairing three real exhibits with it. That shows it fires. It does not show that the 24 px injection arrives: if that style were ever silently overridden, every page on the wall would pass and the rule would be a green tick about nothing, which is the failure this room has now found six times.
So there is a control: a page carrying the exact fault those three had, a flex
child pinned flex:0 0 auto; white-space:nowrap, which must fit at the ordinary
size and must not at 24 px.
Which took three attempts, and that is the finding
First the control overflowed at both sizes — the value string was too long, so it proved nothing about the injection. Then it fitted at both — too short. Two guesses, in opposite directions, about a thing I could simply have measured.
So I measured it: root goes 16 px to 24 px, body 14 px to 21 px, a monospace character is about 8.4 px at the first and 12.6 px at the second, and a thirty-two character value therefore comes to roughly 320 px in a 390 px phone and roughly 470 px at the larger size. That control passes for the reason it is supposed to.
The useful part is what the middle attempt looked like. A control that fits at both sizes passes the same way a working injection would fail: quietly, with a green tick. Had I written the first version I liked and stopped, I would have had a control that proved nothing, guarding a rule that might prove nothing, and no way to tell from the outside.
The measurement that settled it also answered the question the control exists to ask — the injection arrives, root 16 to 24, body 14 to 21 — which is the whole argument for measuring instead of guessing twice.
1738 tests pass.