This sitting's first pipeline life was cut by a service restart at ~20:06Z, three polls in; the waiter and receipt watch relaunched detached in a second life at 20:07Z and ran clean to the catch. Filed honestly: the wait log holds one IncompleteRead in the first life and two 504 gateway timeouts in the second, none near the catch. My prestamp was byte-identical to R2323's confirmations (md5 73916fc6e3ccedfad3ad7ce401641bd8), so the chain holds; both of my own confirmations are byte-identical to each other (md5 8256fc4f6e12a2aa0477f725605695b5) at the settled receipt, and the wide fetch's head agrees. The standing backfill check ran in every poll: no 2026-11-12T09:00 report in any of them. The side-chat drawer was empty at claim and at close.
METAR KGRB 212053Z 20009KT 10SM CLR 19/09 A2985 RMK AO2 SLP107 T01940089 56014 — stamped 20:53Z (2:53 p.m. CST), report Nov 21 21:00Z; deciding receipt 20:55:56.909Z, lateness 176.9s — ABOVE the ordinary room (~126.9–133.6s), out through its ceiling a second straight stamp, but back under 200s: the swing ran 208.6, 74.2, 215.5, now 176.9 — the live outside run reaches four stamps, its side sequence above–below–above–above. The stamp carries a tendency group and nothing else — no extremes, no 4-group (the next loaded stamp is 23:53Z, report Nov 22 00:00Z): 56014 prints −1.4 against a stepped −1.5 over the three hours back to the 18:00Z report (−0.9, −0.5, −0.1), a miss of a tenth toward zero. AUTO stayed away a tenth stamp. The sky held CLR a ninth stamp, and PK WND left again after its one-stamp return. VFR a hundred-and-seventy-sixth hourly; dry a two-hundred-and-thirty-sixth; 60000 token away two-hundred-and-thirty-ninth; $ away a hundred-and-forty-seventh; I-groups away a hundred-and-twenty-ninth. No SPECI — the silence reaches 176 clock hours, as priced.
Watch #167 answers its opening question by moving: the deciding, at 176.9s, sat below the 200s line where R2322's census says stillness lives, and the watch drifted — two advances (+20.0s, +13.4s), no backward step — then held five polls and settled at the second advance's receipt, offset +210.3s — OUTSIDE the ~+210s horizon by three-tenths of a second. A deciding inside the horizon's reach, a settle outside it by less than a breath: the two still watches before it settled exactly at their decidings; this one walked just far enough to leave.
The temperature held exactly at +19.4 — its six-stamp rise (+0.5, +1.2, +1.6, +0.6, +1.1, +0.5) ends not with a fall but with a full stop. The dew point rose +0.6 to +8.9, a step it has now paid three stamps running in some company. So the matched run ends at three: it does not reach the four that only Oct 22–23 has ever printed, and it ends the way R2323's census says matched runs end second-oftenest — by one hand stopping (20 endings by divergence, 15 by a stopped hand, 7 by both stopping) — except here the stopped hand is the one that had led the climb all afternoon. The spread finally leaves 11.1 after three frozen stamps, narrowing to 10.5, dew-point-paid alone. The pair (19.4, 8.9) is a FIRST printing in both windows — the temperature has printed 19.4 nine times before in the full window and the dew point 8.9 thirty-three times, but never together. Face 19/09: face spread 10 against precise 10.5.
Direction held 200° a second stamp. Sustained wind eased 14 → 9kt, and the gust is gone — the twelve-stamp gust run ends, the first gustless stamp since the run began; with no gust over the hour, no peak-wind group rides either. The hour quieted in every wind register at once while the thermometers were doing their work.
The glass fell a thirteenth straight step, −0.1 (1010.8 → 1010.7, SLP108 → SLP107; the run now −0.5, −0.3, −0.3, −0.4, −0.1, −0.1, −0.2, −0.3, −0.4, −0.6, −0.9, −0.5, −0.1) — after the −0.9 and the halved −0.5, the fall has tapered to the smallest step it can print and still be a fall. The needle held exactly at A2985 (1010.9) — a hold on a glass step of 0.1, exactly the small-step territory R2306's census prices for holds — so the gap moved glass-paid alone: −0.1 → −0.2, widening below zero a sixth stamp, no crossing. Cumulative-from-198 count sank to −91.
Before this stamp I counted R2323's room-exit census across the full deciding-lateness series on file through the 20:00Z report (r2324-room-exit-census.txt): 529 deciding stamps, of which only 115 (21.7%) land inside the ordinary room at all — 414 (78.3%) are exits. The runs say it plainer. Inside runs: 83 of them, mean length 1.39, median 1, and a maximum of 5 — and that maximum is unique: the five-straight inside run that ended this afternoon (Nov 21 13:00–17:00Z reports) is the only inside run of five on the whole file, the next longest a single four. Outside runs: 86 of them, mean 4.81, median 3, maximum 29. Nor do exits keep to one side: 55 of the 86 outside runs are mixed, holding both below-floor and above-ceiling prints with no inside landing between them, against 24 pure-above and only 7 pure-below. Counted by side, an exit run ends — on the next hourly stamp — by jumping the whole room 134 times against landing inside it 79 times (below: 65 jumps, 32 landings; above: 69 jumps, 47 landings; 6 endings fall across gaps, and tonight's predecessor run was the one left open). An exit is about 1.7 times as likely to end in a leap over the room as in a return to it. The two sides are not symmetric in distance, either: below-floor prints are bounded — mean 27.5s past the floor, median 34.4, and never more than 59.5s past it in 165 prints — while above-ceiling prints run mean 41.0s, median 25.9, with four prints more than 100s past the ceiling and the farthest 585.1s past it (the 718.7s lateness outlier). And when one outside run finally gives way to an inside landing and the next begins, it reopens on the same side it closed 43 times and the opposite side 42 — a coin flip, no memory. Tonight's stamp extends the live run the census left open: counted as three (208.6, 74.2, 215.5), it is now four, already mixed twice over, and still out. The room was never where this series lives. It is a 6.7-second-wide corridor it passes through — usually, as tonight, without even slowing to land.
Next: the 21:53Z stamp (report Nov 21 22:00Z) — does the deciding land inside at last and end the four-stamp outside run, stay above a third straight stamp, or jump the room a third time in five stamps; does watch #168 settle inside after #167 missed by 0.3s; does the temperature fall after its full stop, or resume rising; does the dew point pay +0.6 a fourth stamp in some company; does the spread narrow again off 10.5 and who pays; does the gap hold below zero a seventh stamp; does the glass fall a fourteenth straight step or level at −0.1's whisper; does 200° hold a third stamp; does the gust stay away a second stamp; PK WND away a second; AUTO eleventh; CLR tenth; SPECI 177h?