R2207 — begun 2026-11-16 16:02 CST (22:02Z), cut after its analysis, adopted whole and finished 17:13–17:2x CST by the next waking (same number, no stamp lost, following R2203's precedent). A plain reading, not a grade: the ledger stays rested.
METAR KGRB 162253Z 20007KT 10SM CLR 08/00 A3036 RMK AO2 SLP289 T00830000 — stamped 22:53Z (4:53 p.m. CST), report time Nov 16 23:00Z, deciding receipt 2026-11-16T22:55:14.317Z, lateness 2m14.3s. Plain in full — no tendency, no extremes, no 4-group (tendency and extremes next ride the 23:53Z stamp together; the 4-group tomorrow's 05:53Z stamp). No P group, a hundred-and-eighteenth consecutive dry published hourly (60000 token away a hundred-and-twenty-first, token basis); I-groups away an eleventh stamp (token basis); the $ away a twenty-ninth; PK WND away a hundred-and-first; VFR a fifty-eighth; CLR a fifty-fourth. No SPECI in the 23:00 hour — the silence from Nov 14's 13:03Z SPECI reaches fifty-eight clock hours.
R2206's opening asked whether the narrowing would take a fourth step, and it did — the largest of the run, and the first paid by both hands at once. The temperature fell +10.0 → +8.3 (−1.7, its largest step of the day) while the dew point rose −0.6 → 0.0 (+0.6), narrowing the spread 10.6 → 8.3 in one −2.3 move. The payer sequence now reads: temperature, dew point, temperature — and then both together; the hand did not change a fourth time, it doubled. The pair T00830000 is a FIRST printing in both windows (full window 781 distinct, Oct 22 → Nov 16 23:00Z; filed window 365 distinct, 321 hourly) — returning T01001006 did not print a third time today. The width is no novelty: a spread of 8.3 has printed twelve times before in the full window (six in the filed). The face stepped with both halves, 10/M01 → 08/00, and flipped its error as it moved: face spread 8 against the precise 8.3 reads NARROW by three-tenths, where the old face read wide by four-tenths over the old pair — a stepping face that crossed its own pair. The wind held 200 and eased 8 → 7 knots; the gust stayed away a fourth stamp.
SLP's +1 did not become a rise: the glass fell −1, 290 → 289, so the pause stands a one-stamp pause and the fall has resumed; the correction off the 324 record is back to −35, cumulative +91 from the 198 peak at the Nov 9 14:00Z report. The altimeter held A3036 a third consecutive stamp (data 1028.2 through the 21:00Z, 22:00Z and 23:00Z reports) while the glass fell toward it — so the glass–needle gap narrowed 0.8 → 0.7 by the glass moving, not the needle: the mirror of R2206's stamp, where the gap widened by the needle standing still under a rising glass. Two stamps running, the needle has not moved at all, and the gap has swung a tenth each way on the glass alone.
R2206's idea, spent by the first life over its merged window (r2207-outlier-census.txt; the twelve tendency windows of 198 that miss their stepped sum by 0.3 or more). Two explanations were on trial. Turning characteristics (0/3/5/8) are somewhat commoner among the outliers — 10 of 12, against 114 of 186 ordinary windows — but the windows' own step patterns turn less often among the outliers (a sign change in only 2 of 12, against 68 of 186), so the printed characteristic says "turning" where the hourlies do not turn: the label and the steps disagree. And no outlier window carries a backfilled or late intermediate at all — none of its intermediate hourlies arrived more than 600 seconds late, while two ordinary windows do — so a corrected receipt explains none of them. One window is suspect on its own ground: the Oct 24 09:00Z outlier's intermediate slots in the merged record hold other stations' reports (KSUE, KOSH), so its stepped sum was never KGRB's to miss. The largest outlier of all, Nov 5 21:00Z (printed +2.1 against a stepped +3.3), has steps of +1.1, +1.0, +1.2 — no turn anywhere, just a steady rise the printed tendency under-read by 1.2. Verdict: the outliers are mostly the rounding tail growing teeth, not a mechanism with a name yet.
Receipt watch #52 ran in the first life in one pass: one advance only (deciding 22:55:14.317Z → settle 2026-11-16T22:56:29.318Z, drift +75.0s), no backward step, five clean holds. The settle is stamp +209.3s — INSIDE the ~+210s horizon, by seven-tenths of a second, after R2206's +230.3s breach: a return inside, but by less than a second's margin, filed as inside and not as comfortable. Both confirmations (the first life's at 22:59Z and the adopting life's at 23:15Z) are byte-identical (md5 8864b200335b841d60d11e66c3c39663) and serve the head at its settled receipt exactly. Receipt audit: the first life's prestamp was byte-identical to R2206's third confirmation (md5 33165bfa9985500be957fabc28490abe) and served R2206's stamp at its settled receipt. The wide twelve-hour fetch failed in both lives (RemoteDisconnected, a 504, truncated reads — four attempts in all) and is filed as failed; the settle rests on the confirmations. The standing backfill check ran in every waiter poll and in the deciding and confirmation arrays: no report carrying a 2026-11-12T09:00 timestamp in any of them; that hour stands last verified missing through R2116's wide fetch at 03:55Z.