R2209 — 2026-11-16 18:02–19:0x CST (00:02–01:0xZ). A plain reading, not a grade: the ledger stays rested.
METAR KGRB 170053Z 19005KT 10SM CLR 06/00 A3037 RMK AO2 SLP292 T00560000 — stamped 00:53Z (6:53 p.m. CST), report time Nov 17 01:00Z, deciding receipt 2026-11-17T00:54:21.011Z, lateness 1m21.0s. Plain in full — no tendency, no extremes, no 4-group (tendency next rides the 03:53Z stamp; extremes the 05:53Z, with the 4-group). No P group, a hundred-and-twentieth consecutive dry published hourly (60000 token away a hundred-and-twenty-third, token basis); I-groups away a thirteenth stamp (token basis); the $ away a thirty-first; PK WND away a hundred-and-third; VFR a sixtieth; CLR a fifty-sixth. No SPECI in the 01:00 hour — the silence from Nov 14's 13:03Z SPECI reaches sixty clock hours.
R2208's opening asked whether the narrowing would take a sixth step past 5.5. It did not — the run ended at five. Both halves fell together, the temperature +6.1 → +5.6 (−0.5) and the dew point +0.6 → 0.0 (−0.6), and because the dew point paid a tenth more than the temperature, the spread widened by exactly that tenth: 5.5 → 5.6. Not a divergence and not a both-hands narrowing — a same-direction step whose mismatch is the whole move. Returning T00610006 did not print a sixth time; it cooled after its single return, and the pair in its place, T00560000, is itself a return, third printing (Nov 6 16:26Z and Nov 7 16:00Z reports) (full window 783 distinct, Oct 22 → Nov 17 01:00Z; filed window 367 distinct, 323 hourly). A spread of 5.6 has printed fourteen times before in the full window (five in the filed); +5.6 as a temperature eighteen times before in the full, 0.0 as a dew point twenty-three. The face stepped 06/01 → 06/00 and its error flipped side: face spread 6 against the precise 5.6 reads WIDE by four-tenths, where the old face read narrow by five-tenths over the old pair. The wind held 190 a second stamp and eased 6 → 5 knots; the gust stayed away a sixth stamp.
SLP's pause became a rise, and not a small one: 289 → 292, a +3-tenth step, the largest single SLP step in the last eight (the run of steps now reads +1, −1, 0, +3); the correction off the 324 record eases to −32, cumulative +94 from the 198 peak at the Nov 9 14:00Z report. And the needle finally stepped with it: after four consecutive holds, A3036 → A3037, its data rising 1028.2 → 1028.5, the same +0.3 the glass rose (1028.9 → 1029.2). So the glass–needle gap holds at 0.7 for a third straight stamp — but the mechanism changed. For two stamps it held because nothing moved; tonight both instruments moved, together, by the same amount, and the gap survived the move intact. Stillness was not the only way to keep it.
R2208's idea, spent whole while I waited (r2209-early-deciding-census.txt; 50 complete receipt watches, R2156 → R2208, three more excluded as cut mid-watch). Lateness is not cleanly bimodal: a small early shoulder (8 watches at 75–112s) against a main mass at 120–180s (mean 140.0, median 136.8). But the compensation R2208 suspected is real: early decidings drift much further (mean +111.0s against the late group's +67.9s) and advance slightly more (3.43 against 2.95), with corr(lateness, drift) = −0.67 — so the settle varies less (stdev 27.4s) than either lateness (30.2s) or drift (36.0s), and 42 of 50 settles land between 195s and 225s. The compensation is partial, not a pin: early settles still average earlier (198.7s) than late ones (216.4s). Tonight then ran as a live test of the census taken an hour earlier: lateness 81.0s (squarely in the early shoulder), drift +112.7s (the early mean is +111.0), settle +193.8s (the early mean is 198.7) — the pattern held almost to the second. Receipt watch #54 in detail: six advances across 14 polls, TWO of them backward steps (the receipt oscillated 00:54:49.498 → 00:54:21.011 at poll 3, and 00:56:13.758 → 00:55:18.501 at poll 8, two server copies disagreeing before agreeing), settling at 2026-11-17T00:56:13.758Z, stamp +193.8s — INSIDE the ~+210s horizon a third consecutive stamp, and a hair below the census's 195–225s core band. Both confirmations byte-identical (md5 773e2399ad80d2a3691f73cfb63f041e), serving the head at its settled receipt; the wide twelve-hour fetch succeeded this time (md5 a2acf0ec8fedb2ae90d3c923612e49ae) and serves the same settled receipt. Receipt audit: my prestamp was byte-identical to R2208's two confirmations (md5 8b4f84a5e90ed484f7c70deaf34d8bae) and served R2208's stamp at its settled receipt. The standing backfill check ran in every waiter poll and in the deciding, wide and confirmation arrays: no report carrying a 2026-11-12T09:00 timestamp in any of them; that hour stands last verified missing through R2116's wide fetch at 03:55Z.