# Job 2785 (route 127 pursuit): pre-registration, written before any population B is computed Only gate cells (#1267, #1389 gates) were run before this file was hashed. Objects as in #1267: base-P twin-admissible slots (r != 0,-2 mod q for q | P); killed by R if r = 0 or -2 mod some q in R; A = K*(P,R), B = K*(Pp,R), drop = A - B. Instrument: kfork.c (this job). Per cell it also gives dstar = min over maximal A-runs rho and translates t in Z/p of k_t(rho) = #{r in rho: r+tM = 0 or -2 mod p} F = max over all killed runs rho of (|rho| - min_t k_t(rho)) (no-merge floor). Theorems used (not tested): #1267 gives B >= A - dstar >= A - floor(2A/p); hence drop >= 2 only if p <= A. The route-127 next step's per-run identity "removed = #forked" is definitional (base-Pp admissible = base-P admissible minus the p-forked slots), so it is not tested as a hypothesis. Population: P = 30, R from primes 7..97, p prime, p !| 30, p not in R, p <= 2A (hard rows); |R| = 3 and |R| = 5 with M = 30*prod(R) <= 3e7; |R| = 4 with M <= 1e7. Gate: instrument must reproduce #1267 (A=12,B=10) and the #1389 gate values before any population row is kept. Checks: C1 (instrument sanity, a theorem): every row has B >= F and B >= A - dstar. A violation is a bug, not a finding. H1 (falsifiable): on every row with dstar >= 1, B = A - dstar (the drop is exactly the minimum translate fork count of a maximal run; no shorter run and no merge through a removed unkilled slot beats it). Falsified by any row with B > A - dstar and dstar >= 1; record all such rows with (R,p,A,B,dstar,F). H2 (falsifiable): B = F on every row (the no-merge floor is exact). Falsified by any B > F. Report: count of rows with p <= A (drop-capable), and the drop histogram on them.