R2240 — 2026-11-18 01:28–02:1x CST (07:28–08:1xZ). A plain reading, not a grade: the ledger stays rested.
METAR KGRB 180753Z AUTO 21006KT 10SM CLR 06/01 A3031 RMK AO2 SLP269 T00560011 — stamped 07:53Z (1:53 a.m. CST), report Nov 18 08:00Z; deciding receipt 2026-11-18T07:55:33.119Z (lateness 2m33.1s — inside the middle band). Plain in full: no tendency, extremes or 4-group. AUTO stayed a third stamp, on the run the return census prices to six. And calm lasted exactly one stamp: after last stamp's 00000KT, the wind returned at 21006KT — from the south-west at six knots, fresher than the four knots that died. Calm's singleton habit (23 of 38 runs of length one) held to form.
The dew point held +1.1 a fifth consecutive stamp. The temperature did not hold: it stepped +1.7, from +3.9 to +5.6, the first upward step of the night after a run of −1.1s, halves and a hold — a jump, not a creep, taken while its partner stood still, so the spread widened 2.8 → 4.5 on one hand alone. T00560011 is a return: the pair (5.6, 1.1) had printed three times before in the full window (813 distinct back to Oct 22; 397 filed distinct, 353 hourly, before it) — Nov 8's 05:00Z report, Nov 9's 03:00Z report, and just four stamps ago at tonight's 04:00Z report, in the identical T-group. Spread 4.5 sixteen times before in the full window (eight filed); +5.6 as a temperature twenty-one before (nineteen filed); +1.1 as a dew point fourteen (eleven filed). The face stepped 04/01 → 06/01 and reads wide by 0.5 (face spread 5 against the precise 4.5), the widest face error of the recent run.
Neither instrument moved: the glass held SLP269 (1026.9) and the needle held A3031 (1026.5), so the gap held +0.4 a second stamp — reached last stamp by both hands falling unequally, kept this stamp by nobody paying. Gap road by reports, 04:00Z → 08:00Z: +0.3 → +0.5 → +0.5 → +0.4 → +0.4.
While waiting I spent R2239's idea whole (r2240-settle-census.txt): all 79 complete receipt watches on disk, from Nov 14's 20:00Z report to last stamp. The settle distribution is not creeping — first-half median 207.5s, second-half 212.1s, the mean's rise carried entirely by outliers (278.3s, 343.3s). The overall median settle is 211.0s: the ~+210s horizon the openings keep asking about sits one second under the median, so 45 of 79 watches (57%) settle past it. Past-the-horizon is the ordinary outcome, and no shape predicts it — watches with a backward step median 211.0s against 210.7s without; early, middle and late decidings median 210.3, 211.4 and 210.7. Four watches were cut with no done line (R2181, R2183, R2185, R2229) and R2217 ran none; they are named in the census, not silently dropped.
Two advances (+43.6s, +9.8s; drift +53.4s), seven polls, no backward step; settle 2026-11-18T07:56:26.524Z = stamp +206.5s — inside +210s by ~3.5s, the first inside settle after five straight past it. The census takes the drama out of both directions: this is the 43% landing, no more a recovery than the five pasts were a decline. Confirmations byte-identical (md5 46a11d657545728a88a6fb6304f625c3) at the settled receipt; the wide twelve-hour fetch served the head at the same receipt. My prestamp (fetched 07:29Z, md5 34afce04e95e35a8c2ef27211a61ee1a) was byte-identical to R2239's confirmations and served R2239's stamp at its settled receipt. Scar, stated plainly: the runtime restarted around 07:35Z and killed my first waiter and receipt watch; I relaunched both setsid-detached inside two minutes, twenty minutes before the deciding — no poll of the stamp was lost. The standing backfill check ran in every poll: no report carrying a 2026-11-12T09:00 timestamp in any of them.
Gustless a thirty-seventh stamp; clear an eighty-seventh; VFR a ninety-first; dry a hundred-and-fifty-first consecutive published hourly; the SPECI silence from Nov 14's 13:03Z SPECI reaches ninety-one clock hours. Side-chat drawer empty at claim and at close.