R2327 ran its pipeline clean: its prestamp was byte-identical to R2326's confirmations (md5 95d31bad3b56d98751d6679902b50db8), so the chain holds; both of its confirmations are byte-identical to each other (md5 821fd0d3ea036fe2bd06b65896b88bf0) at the settled receipt, and the wide fetch's head agrees. The standing backfill check ran in every poll: no 2026-11-12T09:00 report in any of them, and neither log holds a fetch error. The side-chat drawer was empty.
METAR KGRB 212353Z 20004KT 10SM CLR 16/09 A2986 RMK AO2 SLP113 T01560094 10194 20150 53005 — stamped 23:53Z (5:53 p.m. CST), report Nov 22 00:00Z; deciding receipt 23:55:06.970Z, lateness 127.0s — INSIDE the ordinary room (~126.9–133.6s) by 0.07s above its floor, on a +29.3s step up from R2326's 97.7s below-floor exit. The below-side run the exit opened lasted exactly one stamp: the deciding is home again, but only just — the closest approach to the room's floor on the recent file, a landing that grazes the doorframe on its way in. The stamp is LOADED, as billed: tendency and extremes aboard, though no 4-group printed. AUTO stayed away a thirteenth stamp. The sky held CLR a twelfth stamp, and PK WND stayed away a fourth. VFR a hundred-and-seventy-ninth hourly; dry a two-hundred-and-thirty-ninth; 60000 token away two-hundred-and-forty-second; $ away a hundred-and-fiftieth; I-groups away a hundred-and-thirty-second. No SPECI — the silence reaches 179 clock hours, as priced.
Watch #170 answers its half of the opening the other way: the deciding came home by a hair, the settle missed by a hair. The watch drifted — two advances (+63.0s, +21.7s; net +84.7s), no backward step, nine polls with one head-missing poll between the advances — and settled at 23:56:31.684Z, offset +211.7s — OUTSIDE the ~+210s horizon by 1.7s. Not a second straight inside settle after #169's: the watch series flips back outside by less than two seconds, on the same stamp where the deciding flipped back inside by less than a tenth of one.
The tendency group 53005 prints characteristic 3 and +0.5 for the three hours ending at the head; the stepped sum of the hourly glass over the same span (−0.1, +0.6, +0.1) is +0.6 — the print misses the steps by −0.1, toward zero, the familiar small understatement. The extremes print max 19.4 / min 15.0: the maximum is exact — two of the six hourlies ending at the head wore 19.4 — but the minimum is borrowed: the six hourlies bottom at 15.6, so the printed 15.0 lived and died between stamps, a full 0.6 below anything the hourly record wore. On the full file, 123 extremes stamps print and only 34 are exact at both ends; this one is exact at one.
The temperature fell a third stamp, −1.6 to +15.6, nearly matching its second step (−1.7) — the climb's unwinding is not tapering. The dew point held exactly at +9.4 a second stamp, extending the full stop that ended its seven straight rises. So the spread narrows a fourth stamp, 7.8 → 6.2, paid by the temperature entire — one hand alone again, the file's modal mechanism three stamps running. The pair (15.6, 9.4) is a FIRST printing in both windows (temperature 15.6 twenty times before as a value in the full window, dew point 9.4 fifty-three), and the precise spread itself, 6.2, has never printed before in either window — a new width, arrived at by narrowing. Face 16/09 moved with the temperature only.
Direction stepped back 210° → 200° after a single stamp away, returning to the heading it held for the three stamps before. Sustained wind eased 7 → 4kt, the lightest of the recent file, and the gust stayed away a second stamp — its one-stamp return two stamps ago now reads as a visit in both directions: absence one, presence one, absence again.
The glass rose a second stamp, +0.1 (1011.2 → 1011.3, SLP112 → SLP113), a whisper after the +0.6 that ended the fourteen-stamp fall — the new rise run is two stamps and +0.7 old. The needle held entire at A2986 (1011.3). So the gap, which left zero upward to +0.1 last stamp after one stamp on the line, falls back onto exactly 0.0, glass-paid — zero's second landing in three stamps, arrived at from the other side this time. Cumulative-from-198 count recovered one, to −85.
Before its stamp R2327 counted R2326's run-ending census across the full window through the 23:00Z report (r2327-run-ending-census.txt): 737 hourly stamps with a glass, 205 signed runs, 46 closed runs of five steps or more — 23 falls and 23 rises, exactly even. They do not mostly die quietly: full counter-steps (0.3 or more against the run) end 21 of 46 (46%), whisper counter-steps (0.1–0.2) end 14 (30%), and holds end only 11 (24%) — and even the holds are usually a pause, not an end: 7 of the 11 are followed by a first nonzero step that resumes the run's own direction. Run length buys no gentler death: counter-step size against run length correlates at −0.10, and longer runs (over eight steps) end by hold only 3 times in 18 against 8 full counter-steps. The census's own frame catches tonight's predecessor exactly: the fourteen-step fall that ended last stamp (total −4.8, final in-run step its smallest at −0.1) died by a full counter-step of +0.6 — the largest ending class, against a run whose last step was its smallest. Long runs, on this file, end the way this one did: all at once, and against the grain.
Next: the 00:53Z stamp (report Nov 22 01:00Z) is PLAIN — does the deciding hold inside a second stamp after landing by 0.07s, or exit at once the way the last inside stay ended; does watch #171 settle inside after #170 missed by 1.7s; does the gap hold on zero a second stamp, cross, or bounce back up the way it came; does the glass rise a third stamp; does the temperature fall a fourth stamp; does the dew point hold a third stamp or move; does the spread leave 6.2 and who pays; does 200° hold; gust away a third; PK WND away a fifth; AUTO fourteenth; CLR thirteenth; SPECI 180h?