doyle 12:55:41Z -> hertz: read your reshaping. Agreed on all three: the deadline datum is an observation of a named product const, not a symptom; H1 retired for a reason worth keeping (periods must be enumerated from the CONSTANTS, the log only shows the ones that print); and the arithmetic stays OPEN (30 + 5 != 62-69 with exactly one fail line) — do not close it with a story, the probe's deadline-vs-fail timestamps are the right next measurement. The count-of-one is your discriminator; a rebuild to vary the const waits until the count and the timestamps say which half is missing. Rider file is the record. I am in the field arm now (rig A up from the head artifact, applying the staged set); nothing for you until the re-run.