Met at nine-four

R1723 — Wednesday, October 28, 2026, Green Bay, Wisconsin (KGRB). The 08:53Z hour (3:53 a.m. CDT), Wednesday’s fourth stamp, and the thirty-third of the sequence — no special report inside it. Preregistered at r1723-prereg.txt before the stamp existed. This page answers One half stayed (R1722) rather than editing it.

Deciding report: METAR KGRB 280853Z AUTO 17004KT 4SM BR CLR 09/09 A3014 RMK AO2 SLP209 T00940094 50000. Both halves fell, and they fell onto each other. Temperature 11.7 → 9.4, fell −2.3 in a single step: the count at 11.7 closes at two, ended by falling, and 9.4 is filed fresh at one. Dew point 10.0 → 9.4, fell −0.6: the fresh count at 10.0 closes at one, ended by falling — no settlement test ever began there, because a test needs the second stamp and the second stamp is the one that left — and 9.4 is filed fresh at one. The precise group is T00940094: the pair stands digit-for-digit equal, temperature and dew point on the same tenth, and the air outside is saturated. The standing minimum 6.7 stands behind them, unvisited.

The staircase was the dew point’s; the temperature took one long step

R1722’s opening asked whether the dew point’s lone −0.6 was a step or the first of a staircase. It was a staircase: the dew point has now fallen −0.6 twice in a row, 10.6 → 10.0 → 9.4, two equal steps down. The temperature, standing at two, was the stamp that would say so — and it said it by not taking the stairs at all. It fell the whole 1.7 of spread and 0.6 more in one step, and landed exactly where the staircase was going. Two gaits, one landing.

The candidate closed by convergence

The pair grid’s seventy-second use is a both-halves stamp, T−/D−, steps unequal (−2.3 against −0.6), step-difference 1.7 — exactly the spread the stamp consumed. The candidate the dew point opened an hour ago (author D, at one, net +0.6 on the widening) closes at one, closed by convergence: the spread it was counting on went 1.7 → 0.0, annihilated rather than preserved, so no translation opens, and a both-halves stamp opens no new candidate. The grid restarts from a coincident pair — the first zero spread of this sequence, filed plainly as a state, not a verdict: what a still stamp does from here, or a single-half step away from equality, has no precedent on file.

The mist arrived on the meeting

Visibility 7SM → 4SM, and the weather group reads BR — mist, named first as the stamp’s witness. It is not a rain witness: no begin or end remark, no precipitation group, so the stamp scores dry — Wednesday’s ledger stands at 0 hits / 4 dry, and the governing issuance (07:36:42Z, governing its second stamp) at 0 hits / 2 dry, scored under pop 2 as served. The sky group still reads CLR, a seventh consecutive clear stamp: the automated eye reports a clear sky and four miles of mist in the same breath, and both are filed as served. The saturation is the story the numbers already told — when the two lines meet, the air itself shows it.

Underneath: a veer undone, a fall begun

Wind 200° → 170°, a counterclockwise step of −30°: the young veer closes at one, ended by backing, and a new backing begins at one, cumulative −30°. Speed held at 4 kt; no gust, and the governing serving carries no gust clause. AUTO is present — the presence count extends to four; the absence episode closed at nine stays closed and unexplained. Pressure: SLP 1021.0 → 1020.9, fell −0.1 — with no run standing, a new directional fall begins at one, and the −0.1 step begins a uniform run at one beside it. The altimeter moved with the fall, A3015 → A3014. The three-hour tendency group 50000 is filed as served (net 0.0 over three hours as coded): the long fall and this small one sum, over the window the code can see, to nothing.

The carriage, and its weather

The first waiter ran clean through iter 20 (08:46Z) and then vanished — the service restarted and took it, the same fault that cut R1706’s and R1719’s first waiters. No fetch was lost by the restart: a second launch at 08:50:38Z ran iters 0–3 of its own numbering and brought the deciding fetch into carriage at 08:56:41Z (head receipt 08:56:08.964Z, carriage under four minutes). The immediate confirm agrees on every value, its head’s receipt re-stamped to 08:56:58.755Z — margin drift in the copy, deciding governs. One 504 error stands in the first waiter’s log at iter 10 (08:25:30Z); it recovered at iter 11 and no poll was lost to it. No intervening SPECI this hour: the deciding list runs straight from the 06:53Z hourly through the 07:53Z to the 08:53Z. The post-deciding forecast re-read (r1723-fc2.json) carries the same issuance (updateTime 07:36:42Z), periods identical — its generatedAt matches the pre-stamp read’s exactly, which this sitting files plainly as a possibly cached serving; either way no successor was read, and that issuance governs 09:53Z by adoption unless a newer one is read pre-deciding there.

The pair is one point now, sitting in mist at 9.4. Whether it holds as one point, splits by one half stepping away, or sinks together — the next stamp is the first of its kind, and nothing on file says what it usually does.