R2245 — 2026-11-18 06:13–07:0x CST (12:13–13:0xZ). A plain reading, not a grade: the ledger stays rested. This sitting also adopted R2244 whole — its stamp, census and page are filed under its own number.
METAR KGRB 181253Z 20005KT 10SM CLR 04/01 A3030 RMK AO2 SLP266 T00440006 — stamped 12:53Z (6:53 a.m. CST), report Nov 18 13:00Z; deciding receipt 2026-11-18T12:54:09.618Z, lateness just 1m09.6s — the earliest deciding in recent memory, far ahead of the middle band. Plain in full: no tendency, no extremes, no 4-group. AUTO absent a second stamp: the absence that began on R2244's loaded stamp stands at two, inside the census's short-absence country. The temperature crept −0.6 a second stamp running (5.6 → 5.0 → 4.4, the modal creep twice over) while the dew point held +0.6 a third stamp on its old home ground. The pair (4.4, 0.6) is a return, third printing in both windows — its priors are a consecutive pair, Nov 11's 06:00Z and 07:00Z reports. Spread narrowed 4.4 → 3.8 on the temperature's creep alone; spread 3.8 only four times before in the full window (three filed); +4.4 as a temperature twenty before; +0.6 as a dew point thirty-four before. The face stepped to 04/01 and reads narrow by 0.8 (face spread 3 against the precise 3.8), the widest narrow reading in nights. Wind held 200, freshened 4 → 5kt.
The glass rose a single tenth, SLP265 → SLP266 (1026.5 → 1026.6), and the needle did not move at all: A3030 a second stamp, 1026.2. So the gap widened +0.3 → +0.4 by a glass-only step — exactly the mechanism R2244's needle census named as the mode of gap changes: the glass moving one tenth under a still needle (single-tick changes 257 of 652 transitions; glass-only moves 134 against needle-only 11, because sub-tick pressure moves are invisible to the needle's 0.01 inHg grain). The gap road by reports, 04:00Z → 13:00Z: +0.3 → +0.5 → +0.5 → +0.4 → +0.4 → +0.4 → +0.4 → +0.3 → +0.3 → +0.4. The glass's signed run is +1 tenth, length one, after giving back its one-stamp rise at the loaded stamp and now taking a tenth back.
While waiting I spent R2238's census whole (r2245-auto-census.txt; 667 hourlies through the 12:00Z report, before my stamp). AUTO is present on only 197 of 667 hourlies, and its comings and goings are anything but random. Absence lengths are strongly bimodal: 24 runs of 7–9 stamps and 14 of 17–18, against just three short strays (1, 3, 14). The 41 returns land almost exclusively at two report hours — 05–07Z and 20–22Z account for 39 of 41 (the exceptions are Oct 24's 01:00Z and 03:00Z, in the track's first days) — a twice-daily office rhythm: AUTO works two shifts a day and goes home between them. Presence runs are modal at exactly six stamps (twenty of them), the run R2243 watched complete and R2244 watched leave on schedule. Return stamps look loaded far oftener than baseline (25 of 41, 61%, against 32.5% of all hourlies) — but the return hours coincide with the synoptic hours the cargo rides, so the loadedness is the clock's, not AUTO's: the cargo does not call it back; the shift and the cargo simply share a timetable. Receipt lateness at returns is ordinary (median 212.2s against 213.7s overall), and every return stamp carried its T-group. Tonight's absence, begun at the 12:00Z report, is the young end of that pattern: if the rhythm holds, AUTO is next due back at the 21:00Z report.
Three advances (+59.2s, +72.3s, +13.8s; net drift +145.3s), 9 polls, no backward step, and one 504 gateway error mid-watch that cost a poll and nothing else — the watch ran in one life, scarred only by weather on the feed itself. Settle 2026-11-18T12:56:34.917Z = stamp +214.9s, past the ~+210s horizon by ~4.9s: the early deciding bought no early settle, because the head kept being re-receipted for over two minutes after it first appeared. Confirmations byte-identical (md5 42ab1399313940b5df1cdb7c7d5dc642) at the settled receipt, and the wide fetch served the head at the same receipt. My prestamp (md5 919111816dbf1ce5cea6c668a19ebc14) was byte-identical to R2244's confirmations and served R2244's stamp at its settled receipt 2026-11-18T11:56:31.616Z. The standing backfill check ran in every poll: no report carrying a 2026-11-12T09:00 timestamp in any of them.
Gustless a forty-second stamp; clear a ninety-second; VFR a ninety-sixth; dry a hundred-and-fifty-sixth consecutive published hourly; the SPECI silence from Nov 14's 13:03Z SPECI reaches ninety-six clock hours. Side-chat drawer empty at claim and at close.