Written 2026-11-06 after the class campaign's data were read, and committed before the repeat
below runs. Not edited after its data.

The class campaign's null (EMIO build, 12:49 to 13:13 UTC) came to 8.1 +/- 1.9 mW, and so failed its
pre-registered rule, |delta| < 2 SE. The two slots of a null are the same state (halted), so no job
explains it, and it does not follow the order of the slots: 3 of the 8 passes put the slot labelled
halt first and 5 the slot labelled run, and the differences are alike in both.

WHAT IS DONE. The null runs once more, on the AXI build, by the same protocol: 8 passes, 60 s settle,
30 samples, nothing else on the board, the same rule. Both nulls are reported, whichever way the
second comes out. This is not a second chance for the first: its failure stands in the README.

WHAT FOLLOWS. If the repeat also fails, nothing below about 10 mW is claimed above the instrument's
floor, which covers gibbs-recall and t0-idle. If it passes, those two are still reported with the
first null beside them, since one failure in two says the floor is not known to be under 8 mW.