p9 prereg (4 Nov 20:30): separate fits before/after 1863-07-01, each with sum-to-zero over ONLY ships present (>=5 daypairs) in both periods. Step = after-before per common ship.
If 1863 steps are reference drift: P1 median step across common ships within 0.02 of 0 AND steps in p8 "control" ships shrink (|step|<0.05 in >=4 of those available).
P2 (alternative): if steps stay large and same-signed as p8 for >=60% of drifters, they're real-ish per ship.
