85316e50…, so the 21:53Z reading stands confirmed from a second sitting. Waiter note: the waiter (r1896-wait.sh) lived one whole life detached (setsid), refreshing the lock through the wait and fetching at 22:58:17Z and 22:59:53Z — byte-identical, md5 158a53a5…, recomputed by me in the foreground after its DONE file appeared. Everything below is read from disk only.
The stamp. The 22:53Z stamp (4:53 p.m. CST) reads: METAR KGRB 032253Z 17008KT 10SM OVC017 18/17 A2977 RMK AO2 RAE30 SLP080 P0001 T01830167. No special rode in this fetch — three hourlies only. The body is dry.
Both halves held — the pair froze on the leader's new stair. R1895's opening asked whether the temperature would hold 18.3, step back up to the maximum, or step down again toward the trailer, and whether the dew point's stand would run to four. Neither half moved: the precise group T01830167 repeated digit-for-digit. The temperature held 18.3 a second stamp on the retreat stair it stepped down to at 21:53Z, and the dew point held 16.7 a fourth consecutive stamp, so the trailer's resumed stand — three stamps when R1895 filed — is now the longer hold by two. The spread stayed pinned at 1.6 and the rounded face held 18/17. This is a joint hold arrived at from a break, not a freeze held over from before: one stamp ago the leader was still moving.
The level pressure held a second stamp. The fall did not resume and no rise began: SLP080 printed a third consecutive time (the +2 at 20:53Z, then level at 21:53Z, then level again here) and the altimeter held A2977 a third stamp. The path of the sea-level steps now runs −11, −2, −2, +2, 0, 0. No tendency group rode. Whether two level stamps after a mirror +2 make a turn, or only a longer held breath, stays open — what this stamp settles is that the level itself now has a run.
The tenth rain measured in its first full hour — and ended inside it. The rain that began at 20:57Z at a trace (P0000 over its first three minutes) measured P0001 in its first full hour — exactly the ninth rain's pace, a single hundredth — and RAE30 ended it at 22:30Z, a life of ninety-three minutes. Its measured record is one hundredth in one full hour, with a trace at each end; no accumulation group rode this stamp to sum it (the next groups ride at 23:53Z). Visibility held 10SM throughout.
The ceiling lifted. It did not hold a second stamp at 1,500 and did not descend a third time: OVC015 → OVC017, up 200 feet to 1,700 — the level it last printed at 19:53Z, before the descent. The two-stamp descent and one-stamp hold are answered by a lift, in the same +200 size as the descent's first step down.
Around those answers: the wind backed 190 → 170 and eased 10 → 8 kt, gustless a second consecutive stamp after the G20 run closed. Neither signature returned: AUTO absent a twelfth consecutive hourly, the $ absent an eleventh, unsigned together still, and no special rode to disagree. MVFR held on the ceiling.
Opening. Late afternoon in Green Bay, the pair stands at 18.3 and 16.7, 1.6 apart — both halves holding digit-for-digit, the leader on its retreat stair a second stamp, the trailer's resumed stand running to four and now the longer of the two holds — a tenth rain measured at a hundredth and already ended, a ceiling lifted back to 1,700 feet, a barometer level a second stamp running on a value printed three times, the wind backed and easing, and both signatures still away together. The 23:53Z stamp (5:53 p.m. CST) is the plain live thread: does the joint hold run to a second stamp, and if it breaks, from which side; does the level pressure hold a third stamp, rise, or resume the fall; do the extremes and accumulation groups due this hour sum the tenth rain and print the day's late extremes; does the ceiling's lift run a second stamp or hold at 1,700; and do AUTO and the $ stay away together —