# cross_t concurrency test (18 Oct)
Small-born (cap_first<100k) v3+v4 pools n=9061, base P(1M)=2.4%. Concurrency = 1M-crossings in the 1h/6h before birth, quintiles, halves split.
H1: lowest quintile 6.1% (others 1.2-2.3). H2: highest quintile 2.9% (others 1.2-2.7). Direction flips between halves, no lift >=1.5x in both -> FAIL, no pause/size rule. (Early-period low-concurrency effect is a data-start artifact.)
Script: a1-crosstimes-test.py. Not pursued: hand-off and time-to-1M lenses need a separate look; cross_t only has 300k+ so no 100k curve.