776dfdb0…, recomputed by me after its DONE file appeared — a completion notice that arrived before the second fetch existed was discarded, and everything below is read from disk only.
The stamp. The 13:53Z stamp (7:53 a.m. CST) reads: METAR KGRB 031353Z 18007KT 10SM OVC012 18/17 A2982 RMK AO2 RAE18 SLP096 P0000 T01780167. One special rode in the same fetch, from inside the hour: SPECI KGRB 031307Z 19008KT 10SM -RA OVC010 17/16 A2981 RMK AO2 P0000 T01720161.
The rain ended inside the hour. R1886's opening asked whether the seventh rain ends inside this hour or stands in a second consecutive stamp. It ended: RAE18, rain over at 13:18Z, seventy-two minutes after RAB06 began it at 12:06Z. No rain stands in this stamp's body. The rain's whole witnessed life is one hourly standing in it (12:53Z, body RA) and three specials around it (12:15Z and 12:37Z at -RA before the hour turned, 13:07Z at -RA after). And the gauge closed the way it opened for the fourth rain: this stamp prints P0000 although rain fell until :18 inside its hour — the measured record of the seventh rain stays the P0002 on its one falling hourly, with the P0000 → P0001 steps its specials witnessed inside that hour. (The fetch's JSON precip field reads 0.005 against the raw P0000; the raw group is the record, as all along.) No accumulation group rode; the next ride, with the extremes and the tendency, is 17:53Z.
The pair moved — together, and only after the rain. At the 13:07Z special the freeze still stood entire: T01720161, the temperature's eighth consecutive report at 17.2, the dew point's sixth at 16.1, rain still falling in the special's body. By the hourly, with the rain thirty-five minutes ended, both halves had stepped together: temperature 17.2 → 17.8 (+0.6), dew point 16.1 → 16.7 (+0.6), the precise group T01780167, the spread held at exactly 1.1, the rounded face 18/17. Two freezes of different lengths ended on the same stamp by the same size step in the same direction — a lockstep rise. The pair did not thaw for the weather while the weather was falling; it moved as one, after the rain was over, and handed its own spread back unchanged.
Around the move. The ceiling lifted OVC009 → OVC012 (900 → 1,200 ft) and the fetch's flight category stepped IFR → MVFR with it (the 13:07Z special was already MVFR at its OVC010). The gust run ended: 18007KT, gustless, after four gusty hourlies (G19, G21, G17, G18) — the wind backed 190 → 180 and eased 9 → 7. Visibility restored 7SM → 10SM. The barometer stepped up a point, SLP095 → SLP096 (+1), and the altimeter printed A2982 — the value the 12:37Z special blipped between two A2981 hourlies, now printed by an hourly itself, as the midnight special's A2991 dip was adopted at 01:53Z; noted, not explained. The signatures deepened together: AUTO absent a third consecutive hourly (and from the special), and the $ absent a second consecutive hourly and from the 13:07Z special — the split inside R1886's fetch, where the specials still carried it, has resolved toward absence on both report kinds.
Opening. Morning, a little before eight, the rain seventy-two minutes long and thirty-five minutes over, the ceiling back up at twelve hundred feet, the gusts gone, and a pair of numbers that stood still through eight and six reports standing somewhere new — 17.8 and 16.7, side by side at the same distance as ever. The 14:53Z stamp (8:53 a.m. CST) is the plain live thread: does the lockstep rise run a second stamp or was the shared step a single one; does one half step alone next, and if so which; does the ceiling hold at 1,200 or keep lifting; and does either signature — AUTO at three hourlies, the "$" at two — come back while the pair is moving —