Filed honestly: my prestamp (md5 3befc36cfe3478a49e6414e41d5c07bf) is byte-identical to R2299's confirmations, so the custodial chain holds. My detached pipeline shell died with its parent session after the receipt watch completed; the confirmations, merge and analysis were run directly by me after the settle, and every figure here comes from that post-settle run. The wide 14-hour fetch failed all four attempts inside the confirmations (a 502, then three IncompleteReads); a separate retry disconnected once and then carried (md5 01962619e4b5520a2d48e19e67144bf8), serving the head at exactly the settled receipt. The standing backfill check ran in every poll: no report carrying a 2026-11-12T09:00 timestamp in any of them.
METAR KGRB 202053Z 17010KT 10SM CLR 18/08 A3007 RMK AO2 SLP187 T01830078 56034 — stamped 20:53Z (2:53 p.m. CST), report Nov 20 21:00Z; deciding receipt 20:55:11.400Z, lateness 131.4s — INSIDE the ordinary room (~126.9–133.6s), a landing from above on a −32.9s step after four straight stamps above it. TENDENCY stamp — the 5-group rides (56034); no extremes, no 4-group at this report. AUTO absent run 10 — it reached ten, as the opening asked. VFR 152nd hourly; dry 212th; 60000 token away 215th; $ away 123rd; I-group away 105th. No SPECI — the silence reaches 152 clock hours. PK WND away 23.
Watch #143: four advances (+11.5s, +49.8s, +9.0s, +12.1s; drift +82.4s), the settled receipt 20:56:33.830Z held for five holds across eleven polls, no backward step. Settled receipt offset +213.8s — OUTSIDE the ~+210s horizon by ~3.8s, a second straight outside after #142's +225.5s, and the nearer of the two by a wide margin. Confirmations byte-identical (md5 5c5b918fec117243c45db01ad841a822) at the settled receipt.
The temperature held at +18.3. The climb that rose six straight stamps, each step smaller than the last (+6.7, +3.9, +3.3, +2.2, +1.7, +0.5), ended the way the taper census says climbs usually end — not by falling, but by standing still at the top of its own arc, eighteen point three degrees above the 0.0 it leapt from at the 14:00Z report. The dew point fell −0.5, to +7.8, ending its own three-rise run (+2.8, +2.2, +1.1) with a turn. So the spread WIDENED 10.0 → 10.5, dew-point-paid alone — the first widening in three stamps, bought entirely by the moisture stepping back. The pair (18.3, 7.8) is a first printing in both windows (full 875 distinct back to Oct 22; sub 459 distinct, 415 hourly) — the seventh straight first-printing pair. Width 10.5 printed five times before in the full window (three in the sub); +18.3 as a temperature thirteen times full (ten sub); +7.8 as a dew point twenty-one full (seven sub). The face 18/08 printed a second straight stamp, and reads the spread as 10 against a precise 10.5 — narrow by half a degree, the error carried entirely by the dew point's rounding up.
The tendency group printed 56034: signed −3.4 (characteristic 6). The stepped three-hour sum over the window's four reports (1022.0 → 1020.9 → 1019.6 → 1018.7; steps −1.1, −1.3, −0.9) is −3.3. So the print overshoots the stepped fall by a tenth — printed minus stepped −0.1, away from zero — against the tendency census's modal behaviour, which rounds toward zero. A small miss, in the less common direction.
The wind eased 13 → 10kt and veered back 160° → 170°, no gust — the gustless run reaches 3. That −3 is the build's post-peak step: the six-stamp post-calm build (3, 5, 8, 11, 11, 13) has topped, and it eased by an ordinary amount rather than collapsing — 10kt of sustained wind remains, no calm, the direction steady in the 150–170° sector it has held since the return. PK WND's absence reached 23. The sky stayed CLR, its run reaching 3; VFR holds.
The glass fell a seventh straight step at a shrunken size: −0.9 (SLP196 → SLP187, 1019.6 → 1018.7), the fall run now reading −0.3, −0.5, −1.2, −1.4, −1.1, −1.3, −0.9. The needle fell −1.0 beside it (A3010 → A3007, 1019.4 → 1018.4), a tenth harder, so the gap WIDENED +0.2 → +0.3, needle-paid — the two instruments swapping who pays, a stamp after the glass out-fell the needle. The cumulative-from-198 count reaches −11.
Across the 723 hourlies ending Nov 20 20:00Z (source r2299-merged.json), file every maximal non-decreasing sustained-wind run of ≥5 stamps gaining ≥8kt: there are 14, and tonight's is the longest run of any kind in the length table — 14 stamps (eight calm stamps at 0kt, then 3, 5, 8, 11, 11, 13), the sole 14 among non-decreasing increasing runs (next longest 13 and 12, one each). Of the 13 closed qualifying builds, the post-peak step at +1 is negative in all 13 — partly by construction, since a maximal run ends when it falls; the non-tautological content is the size: mean −2.23kt, modal −1 and −2, worst −6, and the peak is never held exactly at +1 (0 of 13) while calm returns within three stamps of the peak only once in 13. Strict builds (≥3 consecutive rises) number 18, the longest five steps (Nov 19, +6kt). So a build's peak is a top, not a plateau — but the descent off it is an easing, not a collapse: the wind keeps most of what it built. Tonight's −3 post-peak step, leaving 10kt standing, is an ordinary member of that distribution, a touch larger than its mean.
Files: r2300-prestamp.json (md5 3befc36cfe3478a49e6414e41d5c07bf, byte-identical to R2299's confirmations), r2300-deciding.json, r2300-wait.log, r2300-receipt-watch.log (four advances, five holds), r2300-confirm.json and r2300-confirm-2.json (md5 5c5b918fec117243c45db01ad841a822), r2300-wide12.json (md5 01962619e4b5520a2d48e19e67144bf8, carried on a post-pipeline retry after four failed attempts and one disconnect), r2300-merged.json (875 distinct stamps, Oct 22 02:00Z → Nov 20 21:00Z), r2300-analysis.txt, r2300-wind-build-census.py/.txt (own devising; source r2299-merged.json, 723 hourlies through the Nov 20 20:00Z report — 14 qualifying builds). Recompute everything: the disk is the record.