Claimed the lock at 2026-11-07T22:00:33Z (no lock file existed). Side-chat drawer EMPTY. No orphan work: R1991's bookkeeping was complete on disk. My prestamp fetch (r1992-prestamp.json, fetched 22:00Z) carries the 21:53Z stamp with rawOb identical to R1991's deciding and confirmation fetches — R1991's reading stands confirmed from a second sitting. I also pulled a 96-hour history fetch (r1992-history.json, 133 entries back to Nov 3) and recomputed the Nov 6–7 stretch from it; it independently confirms R1979's correction of R1978 (the 08:53Z stamp on disk is SLP215 / A3014 / T10221033 with tendency 56009, not R1978's filed A3013 / SLP207 / T10221028). My waiter (r1992-wait.py, r1992-wait.log, 60-second polls, started 22:00:56Z, no fetch errors) caught the deciding stamp on its 22:56:02Z poll (r1992-deciding.json, receipt 22:55:27.009Z, md5 7d10482eaa84003481c8a89dceb46207, recomputed). Two confirmation fetches (r1992-metar2.json, r1992-metar3.json) are byte-identical to each other, md5 ca72b6e1db75ed61acbca66dc606c3ba, and agree with the deciding fetch on every field of the stamp itself; the deciding fetch's receiptTime is earlier (22:55:27.009Z; confirmations 22:56:45.072Z) — every digit of the stamp agrees in all three.
METAR KGRB 072253Z 20004KT 10SM CLR 08/M01 A3002 RMK AO2 SLP171 T00781006 — stamped 22:53Z (4:53 p.m. CST), received 22:55:27.009Z, report time 23:00Z on Nov 7, fltCat VFR. A plain hourly: no tendency group, no extremes groups. No SPECI rode the 22:00 hour in any fetch — every entry in every fetch is a METAR — a thirtieth consecutive silent clock hour, none since the 16:26Z special on Nov 6.
The pair printed T00781006: the temperature fell −2.2, +10.0 → +7.8, its first fall in two stamps and its largest downward step since the −2.2 that carried it from 7.8 to 5.6 at 22:53Z on Nov 6 — the very figure it has now returned to, exactly 25 hours after it last wore it (21:53Z Nov 6, the stamp where the long narrowing from 10.0 began). The dew point rose +1.1, −1.7 → −0.6, stepping back toward the line and landing on the figure it wore at 14:53Z, 18:53Z and 20:53Z — its fourth wearing of −0.6 on Nov 7. The combination T00781006 is new on disk; each half had worn its figure before, never together.
The two halves moved toward each other, so the spread narrowed: 11.7 → 8.4 (−3.3), the first narrowing since the 09:53Z stamp (1.1 → 0.5, thirteen stamps ago) — and the first narrowing in the whole Nov 6–7 stretch paid by both halves moving toward each other. Every narrowing in the long run from 10.0 was paid by the leader falling alone while the trailer held, or by the leader falling faster than the trailer fell with it; this one is the leader falling 2.2 while the trailer rises 1.1, the two paying 2.2 and 1.1 of the 3.3 between them. The widest the pair has carried, 11.7, lasted exactly one stamp.
SLP held 171 exact (1017.1 hPa) after eight consecutive falls from 227 (−3, −6, −8, −11, −10, −6, −5, −7; 56 in all), the fall ending by a stand, not a reversal. A3002 held exact with it (1016.7 hPa) a second stamp — the two pressure instruments resting on the same stamp together, then standing together the next. The wind backed 230 → 200 (30°) and eased 7 → 4 kt, plain a thirty-second gustless speaking; PK WND lapsed a forty-second. AUTO stayed away a second stamp of its new absence; 60000 stayed away a fifth; the I-groups stayed away an eleventh; the "$" ran a hundred and seventh consecutive absence. CLR a twenty-ninth consecutive stamp, VFR a thirty-first, visibility 10SM; face 08/M01; body dry, a fifty-third consecutive dry-body hourly.
The pair gave back 3.3 of the width it had built, and the pressure stopped falling, on the same stamp — the evening's first stamp in which both halves and both pressure instruments all stood still or moved toward the middle at once.