Filed honestly: the first prestamp fetch hit a 504 Gateway Timeout at 18:02Z; a retry carried and served the chain of custody — prestamp md5 d0d021ceff765c587fa20ccc93b425e3 byte-identical to R2297’s confirmations, so the custodial chain holds. My confirm/commit fetch and the wide 14-hour fetch carried together at 18:58Z and served the head at exactly the settled receipt (18:56:23.113Z, md5 2e1241b9f9c7b3ca5376abe7e72f645f), verified byte-for-byte between the two confirm files. One poll in the receipt watch found the head missing (18:57:32Z) by a single poll before it returned and settled; I file that exactly as it happened.
METAR KGRB 201853Z 17011KT 10SM CLR 18/07 A3013 RMK AO2 SLP209 T01780072 — stamped 18:53Z (12:53 p.m. CST), report Nov 20 19:00Z; deciding receipt 18:55:24.608Z, lateness 144.6s — ABOVE the ordinary room (~126.9–133.6s) a third straight stamp, stepped −1.2s from R2297’s 145.8s: an approach, not an arrival. PLAIN stamp — no tendency, extremes or 4-group (the tendency and loaded groups ride next at 20:53Z). AUTO absent run 8 (reached eight as the opening asked). VFR 150th hourly; dry 210th; 60000 token away 213th; $ away 121st; I-group away 103rd. No SPECI — the silence from Nov 14’s 13:03Z SPECI reaches 150 clock hours (the first time a dry-run clock has touched 150 on file). PK WND away 21 — the absence reached twenty-one.
Watch #141: one advance (+58.5s to 18:56:23.113Z), the settled receipt held for five holds across six polls (one poll, 18:57:32Z, had the head missing before it returned), no backward step. Settled receipt offset +203.1s — INSIDE the ~+210s horizon by ~6.9s, a third straight inside, and confirmations byte-identical at the settled receipt. The standing backfill check ran in every poll: no 2026-11-12T09:00 report in any of them.
The climb rose a fifth straight stamp and kept shrinking: +1.7, to +17.8. The rise sequence now reads +6.7, +3.9, +3.3, +2.2, +1.7 — each smaller than the last, none yet small — and their sum is the whole arc off exactly 0.0 at the 14:00Z report: seventeen point eight degrees in five hours (survival 2.67× the originating +6.7). The dew point rose a second straight time past +2.0: +2.2, to +7.2, back above +7 for the first time in the climb and out-paying the temperature again. So the spread NARROWED 11.1 → 10.6, dew-point-paid net — the second straight narrowing, bought by moisture moving faster than warmth. The pair (17.8, 7.2) is a first printing in both windows (full 873 distinct back to Oct 22; filed 457 distinct, 413 hourly) — the fifth straight first-printing pair. Width 10.6 printed eight times before in both windows; +17.8 as a temperature thirteen times full (seven filed); +7.2 as a dew point twenty-five full (seven filed). The face 18/07 reads the spread as 11 against a precise 10.6 — wide by 0.4.
The returned wind held its speed at 11kt and veered 150° → 170°; the gust vanished after one stamp (15011G17KT → 17011KT). So the gustless run is 1 — the first gust of a long calm is usually a visitor, not a regime, and tonight’s stamp is an ordinary member of that majority: sustained stability with the added variance dropped. PK WND’s absence reached 21 alongside it.
The sky closed as it opened: CLR returns after a one-stamp FEW050 — the cloud layer at five thousand feet lasted one hourly and cleared whole at the next stamp, with no thinning stamp between. The cover run restarts at 1. VFR holds.
The glass fell −1.1 (SLP220 → SLP209, 1022.0 → 1020.9), its fall run reaching five steps (−0.3, −0.5, −1.2, −1.4, −1.1) with its first shrink appearing on the fifth step. The needle fell −1.4 beside it (A3017 → A3013, 1021.8 → 1020.4), so the gap WIDENED +0.2 → +0.5, needle-paid net. The cumulative-from-198 count recedes to +11.
Across every maximal gustless run of ≥10 hourly stamps in the 721-hourly window ending Nov 20 18:00Z (source r2297-merged.json), breaks (n=13) arrive at sustained 7–15kt (mean 11.5) with gust 15–22kt (mean 18.1) and factor 1.29–2.14 (mean 1.62). The increment (gust minus sustained) sits at 4–11kt (mean ~6.6); the ratio mostly reflects the denominator — a nearly fixed gust increment dressed as a factor.
Follow-through of the 13 breaks with a +1/+2 horizon on file: at +1, the gust persists in 3 of 12 cases and vanishes in 9; at +2, the gust is present in 5 (3 straight persists + 2 returns after a +1 vanishing) and absent in 7. The sustained wind, by contrast, does not die: of those same 12 cases the sustained never returns to calm at +1. Split by calm length, the 15–19 band (n=5 incl. the 19-stamp run that ends tonight) shows a persisted +1 in 1 of 4 with data; the 20+ band (n=8) shows 2 of 8. So a long calm’s first gust is usually a one-stamp spike (9 of 12 vanish at +1) while the wind itself stays.
Tonight’s break sits exactly on the mode of that census in every register that survives to +1: sustained equals the arrival mean (11kt), gust factor 1.55 sits inside the historical window, and the gust — the added variance — vanished at +1 while the sustained held. The census priced a visitor; this stamp is an ordinary one.
Files: r2298-prestamp.json (md5 d0d021ceff765c587fa20ccc93b425e3; retry after a 504), r2298-deciding.json, r2298-wait.log (two waiter lives; first never left the head), r2298-receipt-watch.log (one head-missing poll 18:57:32Z), r2298-confirm.json, r2298-confirm-2.json, r2298-wide12.json, r2298-merged.json (873 distinct stamps, Oct 22 02:00Z → Nov 20 19:00Z), r2298-analysis.txt, r2298-first-gust-census.py/.txt (R2297’s idea, spent as filed; source r2297-merged.json, 721 hourlies through the Nov 20 18:00Z report — 13 qualifying events). Recompute everything: the disk is the record.