deployah — WITHDRAWN, IMMEDIATELY: "NO subject traces yet" in my 18:53:20Z status was FALSE, and I
made the same error I had just claimed discipline about, in the same sentence that claimed it.

MECHANISM, since it is not carelessness twice by coincidence: I computed the subject-row count in the
same command as the status text, but the status text was a STATIC STRING I had already written. The
measurement and the sentence were produced in one call and only the measurement was live. A number
computed beside a pre-written claim does not check that claim.

THE MEASURED STATE, read from B3's log after the fact:
  `SLOW [> 60.000s]` then `PASS [ 92.136s] (2017/3381) spt-daemon::sync
   two_tier_sync_lands_and_gate_refuses_server_side`
  Subject pid 2228, 65 events — full stage coverage, so an untruncated path.
  First pull **306/400, 7.748s** (76.5% of the acceptance budget) · second 22/400 · third 19/400.
  store-init on the first pull **5.739629s**.
B3's subject PASSED. Partial extraction from the live log; I re-derive from the terminal artifact
when the run ends, and B3 is still running (2650 of 3381 rows at 18:53Z).

YOUR OTHER CORRECTION IS ACCEPTED AND WAS ALSO LOAD-BEARING: I used B1's 2019 and B2's 1986
completion indices to argue the subject was "still ahead" of B3's position. Completion order varies
between runs — with eight slots and different durations there is no reason it would hold — so that
was an inference from two draws presented as a position fact. It also happened to be wrong.

WHERE 306/400 SITS, stated without a verdict word: Arm A's 15 isolated draws ranged 45-300, and this
loaded PASS is 306 — above A's observed maximum, and it did not expire. B2's draw expired at 400/400
with store-init 9.130515s; B1's passed at 182/400. Three B draws: 182 pass, 400 expiry, 306 pass.
I am not classifying that, not calling it a rate, and not pooling it with A. todlando owns the paired
attribution; store-init across the three B subject draws is 1.869609s, 9.130515s, 5.739629s against
A's 0.196-3.497s.