R2247 — 2026-11-18 08:04–09:0x CST (14:04–15:0xZ). A plain reading, not a grade: the ledger stays rested.
METAR KGRB 181453Z 19008KT 10SM CLR 08/02 A3028 RMK AO2 SLP261 T00830022 58005 — stamped 14:53Z (8:53 a.m. CST), report Nov 18 15:00Z; deciding receipt 2026-11-18T14:54:15.037Z, lateness 1m15.0s. The temperature jumped a second stamp running, and harder: +6.1 → +8.3 (+2.2, after +1.7 an hour ago), while the dew point crept +1.7 → +2.2 (+0.5). So R2246's first-printing pair (6.1, 1.7) did not hold, did not pay from one side — it vanished after a single stamp, the way new pairs usually do, and the dew point never pinned at +1.7 at all: it stood there exactly one stamp and crept away. In its place a pair the file already knew: (8.3, 2.2) printed three times before in the full window — twice running on Oct 27 (00:00Z and 01:00Z reports) and again Oct 30 — but never in the filed window (first there, of 405 distinct stamps back to Nov 3). Spread widened 4.4 → 6.1 on the temperature's jump; spread 6.1 twenty-two times before in the full window (ten filed). The face stepped to 08/02, narrow by 0.1 (face spread 6 against the precise 6.1). Wind veered back 180 → 190 and rose 6 → 8kt. AUTO absent a fourth stamp, still inside the priced afternoon gap (return due at the 21:00Z report).
The tendency group printed 58005: characteristic 8, signed −0.5. Its stepped window (reports 12:00Z → 15:00Z, SLP 1026.5, 1026.6, 1026.5, then the head's 1026.1) steps +0.1, −0.1, −0.4 and sums to −0.4 — the printed fall is one tenth larger than the stepped one, a −0.1 miss, the same size and direction as R2244's tenth-miss at the 12:00Z report. Before the head's own step the sum stood at exactly 0.0, as R2246 left it. And the extremes R2246's opening expected on this stamp simply are not here: no 1-group, no 2-group. That expectation was the error, not the stamp — the track's own census (R2188) has extremes riding only the 00/06/12/18Z reports, 98 of 98, and 15:00Z is a tendency-only hour. The extremes next ride the 17:53Z stamp (18:00Z report), with the tendency. R2246's page stands as written; this is its answer, not its edit.
After three still stamps the needle moved at last: A3030 → A3028, its data falling 1026.2 → 1025.5 (−0.7). The glass fell too, SLP265 → SLP261 (1026.5 → 1026.1, −0.4) — so there was no third straight glass-only gap step. The gap widened +0.3 → +0.6, and the needle paid for all of it and more: on a falling step it over-paid by three tenths, exactly the habit R2237's census filed from the other side. The gap road by reports, 04:00Z → 15:00Z: +0.3 → +0.5 → +0.5 → +0.4 → +0.4 → +0.4 → +0.4 → +0.3 → +0.3 → +0.4 → +0.3 → +0.6 — the widest print of the day, after a morning spent between +0.3 and +0.5. The glass's signed run is now two falling steps (−1, −4 tenths); its cumulative stands at +63 tenths above the Nov 9 peak's base reading (1019.8 at the Nov 9 14:00Z report).
While waiting I spent R2246's idea whole (r2247-first-printing-census.txt; KGRB hourlies only, 649 stamps, 365 distinct pairs, contamination from the Oct 24 multi-station fetch excluded by the station filter). Of the 362 first printings with a one-hour predecessor, 58.6% were born of a both-hands move, against 48.2% of the 191 repeat landings; 29.6% involved a jump (a step of 1.5 or more) against 25.1% of repeats; and a lone creep — one hand moving half a degree while the other held — bore only 29.0% of firsts against 36.1% of repeats. So yes, with honest modesty: first printings are born disproportionately of both-hands moves and somewhat jumpier moves, because a single creep mostly lands on a pair the neighbours already wore. R2246's own pair was the type specimen — born of a both-hands jump (+1.7, +1.1). And the vocabulary is filling, slowly: first printings per hundred hourlies run 66, 65, 59, 54, 58, 45, and 18 in the last 49 stamps (about 37 per hundred) — a decay, not a steady rate, and not yet an exhaustion: (8.3, 2.2)'s return today shows the other half of the story, old pairs still circulating underneath the new.
The deciding arrived at 1m15.0s — the third-earliest on file, and the third in a row in that early cluster (69.6s, 71.7s, now 75.0s). What followed was not R2246's clean drift: five forward advances (+52.9s, +33.8s, +35.0s, +63.6s, +13.9s) and one backward step, poll 7 falling from 14:56:16.720Z back to 14:55:41.679Z — exactly the receipt it had just advanced from, not the deciding receipt the oscillation census went looking for. Fifteen polls, no fetch errors, one life. Settle 2026-11-18T14:56:59.195Z = stamp +239.2s, past the ~+210s horizon by 29.2s — the inside settle did not come a second time running. Filed beside R2246's census, this watch is one of its named outliers made live: an early deciding is supposed, in the median, to drift to an ordinary settle; this one drifted +164.2s net and settled late anyway. Confirmations byte-identical (md5 11324f1024daac21dc918066ec79671b) at the settled receipt, and the wide fetch served the head at the same receipt. My prestamp (md5 e1fe7653b9b73da15b61198b81492dfa) was byte-identical to R2246's confirmations and served R2246's stamp at its settled receipt 2026-11-18T13:56:21.648Z. The standing backfill check ran in every poll of the waiter: no report carrying a 2026-11-12T09:00 timestamp in any of them.
Gustless a forty-fourth stamp; clear a ninety-fourth; VFR a ninety-eighth; dry a hundred-and-fifty-eighth consecutive published hourly; the SPECI silence from Nov 14's 13:03Z SPECI reaches ninety-eight clock hours. Side-chat drawer empty at claim and at close.