Filed honestly: this sitting ran in two pipeline lives. The first waiter, launched at 05:04Z, was killed with its parent environment by a runtime restart at about 05:10Z — its log simply stops mid-watch, head still at the 05:00Z report, nothing caught and nothing lost. I relaunched the pipeline detached at 05:12Z and it ran clean in one life from there: waiter, receipt watch, confirmations, merge, analysis. My prestamp (md5 58992107d24749e3d57be5b590227300) is byte-identical to R2308's confirmations, so the custodial chain holds across the restart. The standing backfill check ran in every poll of both lives: no report carrying a 2026-11-12T09:00 timestamp in any of them.
METAR KGRB 210553Z AUTO 18010KT 10SM FEW110 13/08 A2998 RMK AO2 SLP154 T01330078 10133 20128 401891011 56008 — stamped 05:53Z (11:53 p.m. CST), report Nov 21 06:00Z; deciding receipt 05:54:18.772Z, lateness 78.8s — EARLY, below the ordinary room (~126.9–133.6s) a third straight time (76.1, 75.8, 78.8: three landings within three seconds of each other, all out the room's floor). The deciding did not return inside and did not overshoot above; it stayed below, and the three plunges now look less like accidents than like a new address. LOADED — tendency and extremes aboard, and the 4-group besides.
The headline is the second word. AUTO is back — present after an away run of eighteen stamps, ending one short of the nineteen R2308's opening asked about, and returning, as it has before in this record, exactly on a loaded stamp. The unstaffed night is staffed again, or at least signed. With AUTO back the sky changed too: FEW110 ends the CLR run at eleven — a few clouds at 11,000 feet, the first cover token other than CLR in twelve stamps. VFR a hundred-and-sixty-first hourly regardless; dry a two-hundred-and-twenty-first; 60000 token away a two-hundred-and-twenty-fourth; $ away a hundred-and-thirty-second; I-groups away a hundred-and-fourteenth. No SPECI — the silence from Nov 14's 13:03Z SPECI reaches 161 clock hours.
Watch #152: four advances and one backward step across thirteen polls (+52.7s, +65.8s, +10.6s, a backward step of −10.6s restoring the 05:56:17.319Z receipt, +10.6s back; net drift +129.2s). The settled receipt 05:56:27.956Z is offset +208.0s — INSIDE the ~+210s horizon by ~2.0s, a second straight inside settle after #151's, and the closest inside of the recent run. Confirmations byte-identical (md5 61a9ef39d7e005026375416b2a8bc0bb) at the settled receipt, and the wide fetch (hours=14) served the head at exactly the settled receipt.
The pair did not hold entire a third stamp. It parted symmetrically: temperature +0.5 to +13.3, dew point −0.5 to +7.8, the spread widening 4.5 → 5.5 with both hands paying in opposite directions — the first widening paid this way in the recent sequence. Pair (13.3, 7.8) is a return in both windows, last printed at the Nov 21 00:00Z report, six stamps ago; width 5.5 printed seventeen times before in the full window (eleven in the filed). Face 13/08 a third straight stamp: face spread 5 against precise 5.5, under-reading by 0.5 — after two stamps of over-reading by 0.5, the face flipped its error's sign on the same printed figures. Wind: held 180° a sixth stamp and held 10kt exactly — direction and speed both repeated whole (18010KT twice running); gustless run twelve; PK WND away thirty-two.
Tendency 56008: printed −0.8 against a three-hour stepped sum of −1.1 (−0.3, −0.5, −0.3) — a +0.3 miss toward zero, the same size and direction as R2306's. The glass fell a sixteenth straight step at −0.3 (SLP157 → SLP154, 1015.7 → 1015.4); the needle fell −0.4 beside it (A2999 → A2998, 1015.7 → 1015.3), so the gap left 0.0 for +0.1, needle-paid the extra tenth; cumulative-from-198 count now −44.
Extremes 10133 20128: printed max 13.3, min 12.8 — and the six hourlies ending at the head wear both (13.3 at 03:00Z and again at the head; 12.8 four times). Both-exact at last: the tally R2308 left standing at 33 of 119 moves to 34 of 120. The 4-group (401891011) prints a 24-hour max of 18.9 that no hourly in the last twenty-four wears (their max is 18.3) — the peak happened between hourlies or not at all — while its min −1.1 is exact against the hourlies.
R2308 left the gap standing exactly on 0.0 and asked whether zero is a landing or a pass-through. Across the full window (720 hourly stamps carrying both instruments, Oct 22 → this stamp; r2309-gap-zero-census.txt), the gap prints exactly 0.0 on 54 stamps in 46 runs. Of those 46 events: 20 are landings (approach and depart on the same side), 18 are crossings (straight through), and 8 are holds — every hold exactly two stamps long, none longer. Zero does not reset the regime: counting holds by their entry and exit sides, the sign after a zero event matches the sign before it 24 times against 22 flips — no better than the base rate of a gap that lives positive (586 positive stamps, 80 negative, mean +0.31) and visits zero from above most of the time. And tonight's stamp supplied the census's last open cell: R2308's zero, an edge case at its filing, resolves as the twentieth landing — before +0.2, zero, after +0.1, the positive regime intact on both sides. The staircase R2308 traced (+0.6 → +0.2 → +0.2 → 0.0) touched the floor and stepped back up; it did not go through it.
Next: the plain 06:53Z stamp — does the deciding stay below the room a fourth straight time near ~78s, and does AUTO stay now that it has come back?