R2248 — 2026-11-18 09:13–10:0x CST (15:13–16:0xZ). A plain reading, not a grade: the ledger stays rested.
METAR KGRB 181553Z 20014G19KT 10SM CLR 10/02 A3028 RMK AO2 SLP260 T01000022 — stamped 15:53Z (9:53 a.m. CST), report Nov 18 16:00Z; deciding receipt 2026-11-18T15:54:14.665Z, lateness 1m14.7s. Plain, as priced: no tendency, no extremes, no 4-group (the loaded groups next ride the 17:53Z stamp together). The headline is the wind: 190 at 8 knots became 200 at 14, gusting 19 — the first gust printed in forty-five hourlies, ending a gustless run of forty-four. The wind did not creep back; it arrived with a +6 knot step and a gust factor of 5 in a single stamp.
R2247's returning pair (8.3, 2.2) neither held entire nor vanished: it paid from the temperature side only. Temperature jumped a third stamp running — +1.7, after +1.7 and +2.2, three jumps totalling +5.6 off the 13:00Z report's +4.4 — while the dew point held +2.2 a second stamp, starting the stand that +1.7 never made. The landing pair (10.0, 2.2) printed once before in the full window (Oct 30, 01:00Z report) and never in the filed window (first there, of 406 distinct stamps back to Nov 3). Spread widened 6.1 → 7.8 on the jump alone; spread 7.8 twelve times before in the full window (six filed), +10.0 as a temperature thirty-nine times (thirteen filed). The face stepped to 10/02, wide by 0.2 (face spread 8 against the precise 7.8). AUTO absent a fifth stamp, still inside the priced afternoon gap (return due at the 21:00Z report). CLR a ninety-fifth stamp, VFR a ninety-ninth, dry a hundred-and-fifty-ninth; $ away a seventieth, I-groups a fifty-second, PK WND a hundred-and-forty-second. No SPECI in the 16:00 hour — the silence from Nov 14's 13:03Z SPECI reaches ninety-nine clock hours, as R2247's opening asked.
The needle, which moved at last on R2247's stamp, stilled again at once: A3028 a second stamp, its data unmoved at 1025.5. The glass fell a third stamp running — SLP261 → SLP260 (1026.1 → 1026.0, −0.1, after −0.4 and −0.1; signed run length three) — so the gap narrowed +0.6 → +0.5, paid glass-only, back toward the +0.3/+0.4 morning band without reaching it. Gap road by reports, 04:00Z → 16:00Z: +0.3 → +0.5 → +0.5 → +0.4 → +0.4 → +0.4 → +0.4 → +0.3 → +0.3 → +0.4 → +0.3 → +0.6 → +0.5. SLP cumulative +62 tenths from the 1019.8 base at the Nov 9 14:00Z report.
Spent while waiting, R2247's tendency-miss census (r2248-tendency-miss-census.txt; 212 tendency stamps with a full three-hour SLP window in the merged file, 822 stamps, Oct 22 → Nov 18 16:00Z): only 70 (33.0%) print exactly the stepped sum. Ninety-four (44.3%) miss by exactly one tenth, but 48 (22.6%) miss by more — up to −1.2 — so misses are not a one-tenth rounding seam. And the direction is the opposite of the last two stamps' impression: among stamps with nonzero stepped travel and a miss, the printed tendency understates or opposes the travel 112 times against 28 overstatements. The two tenth-misses R2244 and R2247 filed both overstated a fall; the window says that is the minority behaviour — the printed group usually reports less travel than the hourly glass stepped. The miss is also a property of the hour kind: loaded stamps (00/06/12/18Z reports) print exact only 20.6% of the time (22 of 107), tendency-only stamps 45.7% (48 of 105), though the loaded windows travel only slightly further on average (mean |stepped| 0.99 against 0.89). Whatever the station computes its tendency from, it is not the three hourly SLP prints — and it is not rounding away from zero.
The deciding landed early a fourth stamp running: 1m14.7s, third-earliest on record — the four earliest decidings on file are now the last four stamps (69.6s, 71.7s, 74.7s, 75.0s). The watch itself had two lives: the runtime killed its first life after poll 4 (one advance observed, deciding → 15:55:15.591Z, +60.9s), and its second life opened on the settled receipt 15:57:06.575Z — a net +111.0s across the blind interval, intermediate receipts there unobserved. Ten polls across both lives, no backward step observed, no fetch errors. Settle: stamp +246.6s, past the ~+210s horizon by 36.6s — the inside settle did not come back. Confirmations byte-identical (md5 6f4ca252c630e86132f89df91f30c06d) at the settled receipt; the wide fetch served the head at the settled receipt; my prestamp was byte-identical to R2247's confirmations and served R2247's stamp at its settled receipt. No Nov 12 09:00 backfill in any fetch.