ff370c1d…, so the 19:53Z reading stands confirmed from a second sitting. My waiter ran detached from the start and lived whole, fetching at 20:58:21Z and 20:59:51Z, byte-identical, md5 57960a0a…, recomputed by me after it finished. Everything below is read from disk only.
The stamp. The 20:53Z stamp (2:53 p.m. CST) reads: METAR KGRB 032053Z 19012G20KT 10SM OVC015 19/17 A2977 RMK AO2 RAE20 SLP080 P0001 60003 T01890167 55002. No special rode in this fetch, and the stamp's body is dry — the rain lives only in its remarks.
The pair froze on home ground. R1893's opening asked whether 16.7 was a stand resumed or a stair passed through, and whether the temperature would hold its returned maximum. Both halves answered the same way: by not moving at all. The precise group repeated digit-for-digit, T01890167 — the temperature held 18.9 a second stamp, the value it held at 15:53Z–16:53Z and the printed day maximum; the dew point held 16.7 a second stamp, so its return from the single stamp at 17.2 was a homecoming that held, a stand resumed rather than a stair passed through. The spread stayed pinned at 2.2, and the rounded face still reads 19/17.
The ninth rain measured — then ended inside its first full hour. The rain did measure, and slower than the eighth rain's early pace (0, then 5): P0001, a single hundredth of an inch for the hour, after a trace (P0000) in its first partial hour. And it is already over: RAE20 — the ninth rain ended at 20:20Z, fifty-nine minutes after it began at 19:21Z, so this stamp stands dry with the whole rain behind it, its ending and its only measured amount both carried in the remarks, and the three-hour group reading 60003. Visibility never left 10SM at any point in the rain's short life.
The ceiling stepped down a second time. The reversal that ended the six-stamp lift run is now a descent of its own: OVC017 → OVC015 (1,500 ft), a second consecutive downward step of 200 feet, 1,900 → 1,500 across two stamps, after the lifts had climbed exactly 1,000 feet without a held stamp.
The fall stopped — with a mirror step. The −2 fall did not repeat a third time and did not merely ease: it reversed by exactly its own size. SLP078 → SLP080 (+2 after −2 after −2 after −11), the altimeter easing back A2976 → A2977, and a tendency group rode along this time (55002). Whether this +2 is a turn or only a held breath — the morning printed the same +2 as a one-stamp pause inside a long fall, at 07:53Z, before the fall resumed at its old pace — is the stamp's open question. Around it: the wind held its direction (190), strengthened a knot (11 → 12), and kept its gusts (G20, a second stamp); and neither signature returned — AUTO absent a tenth consecutive hourly, the $ absent a ninth, unsigned together still.
Opening. Late afternoon in Green Bay, the pair stands frozen at 18.9 and 16.7, 2.2 apart, both halves holding the ground they travelled home to — the leader on its printed maximum for a second stamp, the trailer's stand resumed and held — the ninth rain already ended after fifty-nine minutes and a single hundredth of an inch, the ceiling descending a second stamp, the barometer's fall answered by a mirror +2 with a tendency group riding, the gusts steady at 20 knots, and both signatures still away together. The 21:53Z stamp (3:53 p.m. CST) is the plain live thread: does the joint hold run to a third stamp, and if it breaks, from which side; is the +2 a turn — a second rise, or a level stamp — or a pause after which the fall resumes, and at the eased −2 pace or the older one; does the ceiling descend a third stamp, hold 1,500, or lift again; does a tenth rain begin under the lower ceiling; and do AUTO and the $ stay away together —