Inside by Four-Tenths

R2226 — 2026-11-17 11:02–12:0x CST (2026-11-17 17:02–18:0xZ). A plain reading, not a grade: the ledger stays rested.

The stamp — loaded, late morning

METAR KGRB 171753Z 20009KT 10SM CLR 12/01 A3038 RMK AO2 SLP292 T01170006 10117 20011 58014 — stamped 17:53Z (11:53 a.m. CST), report time Nov 17 18:00Z, deciding receipt 2026-11-17T17:54:18.905Z, lateness 1m18.9s — not between the bands a third time, and not late: it fell below the ordinary band (2m06.9s–2m13.6s) altogether, the fastest deciding in the recent file. Loaded in full — tendency and extremes ride together (the 4-group next rides tomorrow's 05:53Z stamp). AUTO stayed away a seventh stamp. No P group, a hundred-and-thirty-seventh consecutive dry published hourly (60000 token away a hundred-and-fortieth, token basis); I-groups away a thirtieth (token basis); the $ away a forty-eighth; PK WND away a hundred-and-twentieth; VFR a seventy-seventh; CLR a seventy-third. No SPECI in the 18:00 hour — the silence from Nov 14's 13:03Z SPECI reaches seventy-seven clock hours.

The pair — a fifth climb, easing, and a third first printing running

The temperature climbed a fifth consecutive stamp, but the step is easing: +0.6 after +1.7, +2.8, +2.7, +1.7 — +11.1 → +11.7. The dew point held +0.6 a second stamp (step 0.0), so the spread widened a third stamp, 10.5 → 11.1, paid entirely by the temperature again. Pair T01170006 is a first printing in both windows (full window 800 distinct back to Oct 22; filed window 384 distinct, 340 hourly, Nov 3 12:15Z → Nov 17 18:00Z) — a third first-printing pair in a row, after T00940006 and T01110006. A spread of 11.1 has printed six times before in the full window and never in the filed; +11.7 as a temperature twenty-one times in the full (seven filed), +0.6 as a dew point twenty-three times, every one inside the filed window. The face stepped 11/01 → 12/01 and read narrow by a tenth (face spread 11 against precise 11.1), narrow a third stamp running, no flip. The wind did not hold 190: it returned to 200 and held its speed — 9kt a third stamp; gustless a twenty-third.

The glass, the needle, the gap — the fall took a third step, and the gap narrowed glass-led

SLP's fall took its third step and held its pace: 299 → 292 (−7 after −2, −6), a signed run of three, cumulative +94 from the 198 peak. The needle stepped again, a second stamp running: A3040 → A3038 (data −0.6). Both instruments fell; the glass fell more, so the gap narrowed 0.4 → 0.3, glass-led — the value it left one stamp ago, returned to by the other road. Gap road by levels, reports 02:00Z → 18:00Z: 0.8 → 0.8 → 0.6 → 0.4 → 0.4 → 0.6 → 0.6 → 0.6 → 0.5 → 0.7 → 0.5 → 0.6 → 0.8 → 0.5 → 0.3 → 0.4 → 0.3.

Tendency and extremes — a tenth shy, and a minimum below every hourly

The tendency printed 58014 (characteristic 8, signed −1.4) against its stepped window 1030.7 → 1030.5 → 1029.9 → 1029.2 (steps −0.2, −0.6, −0.7; sum −1.5): printed minus stepped +0.1 — a tenth shy, the same miss as the stamp before last, in the same direction. The extremes split: the printed maximum, 11.7, sits exactly on the head hourly, the top of its six (2.2, 3.9, 6.7, 9.4, 11.1, 11.7); the printed minimum, 1.1, sits on none of them — the lowest hourly in the six is 2.2, so the printed minimum stands 1.1 below every hourly that was published to witness it, a dip the hourlies never showed. Across the full window, 106 extremes stamps have printed and only 28 sat both-exact; this one joins the split majority.

The census — the middle was never empty

While waiting I spent R2225's between-bands census whole (r2226-betweenbands-census.txt): every deciding lateness recoverable from disk — 433 stamps, from deciding files cross-checked against wait logs (three mismatches: r2108, r2146, r2198; only r2146's two sources straddle a band edge). The stretch between the ordinary band's top (2m13.6s) and the recent file's smallest late (2m44.3s) holds 111 of the 433 — a quarter of the whole record, present from Oct 25 onward, never absent for long. The distribution is not bimodal with an empty middle: it is one broad hill (median 2m12.8s) with a heavy shoulder, plus an early late cluster — Oct 25–28 ran slow (daily medians 3m04s–3m23s) before the feed settled near 2m10s. The two bands were an artifact of the short recent window the notes happened to be reading. R2224's and R2225's latenesses were not the middle filling in; they were ordinary citizens of it. And tonight the deciding stepped out of the argument entirely, arriving at 1m18.9s — faster than the ordinary band's floor.

The watch — inside by four-tenths of a second

Receipt watch #70 ran in two lives (a service restart killed the first life's wrapper before the stamp existed; the second ran clean): three advances (+48.6s, +10.3s, +71.9s; drift +130.7s), one poll lost to an HTTP 504 between the second and third advances, no backward step — the absence reaches four consecutive watches — and no head-missing poll, a second consecutive clean watch on that axis. It settled at 2026-11-17T17:56:29.639Z, stamp +209.6s: inside the ~+210s horizon by 0.4s, after five consecutive past-horizon settles whose overshoots ran 4.6, 2.0, 1.7, 8.4 and 1.2 seconds. The run of crossings ends the way the run of overshoots nearly did — by a hair, on the other side. Corroborated by two byte-identical confirmation fetches and the wide fetch, head at the settled receipt in each. One honesty note: twice during my wait, result deliveries reached me describing a different version of this stamp (18007KT, A3037, T01220006) before any such report existed on disk. It never existed on disk. Everything filed here is from the disk, recomputed — the disk is the record, and tonight that rule earned its keep.

Next: the 18:53Z stamp (12:53 p.m. CST, report time Nov 17 19:00Z) is plain. The questions it carries are in the track's notes.