v0.441.0 — FRONT 4 RECORD RUN SHIP

B287 (PAPER_2282): the front-4 record run - an engine written for it, four
runs in two boxes with three seeds, the number did not move, the
pre-stated falsifier applied, three rulings folded. One band since v0.440.0.

- THE ENGINE: md_grade/md_pme.py + pair_kernel.c + gridcur.py - rigid
  TIP4P/2005 with smooth particle-mesh Ewald (the reaction-field caveat of
  B286b gone), pair kernel / RATTLE / B-spline spreading in C, transverse
  currents by FFT momentum-density transform. Validated: Madelung constant
  of NaCl 1.747565 (1.2e-7); C vs numpy 1.2e-4 kJ/mol; SPME 3e-3; forces =
  -grad E to 3e-8; NVE -0.0001 kJ/mol per molecule over 2 ps; E/N -47.2;
  eta_0 0.84-0.85 vs 0.855; D 2.3-2.4e-5 vs 2.1-2.3.
- THE RUNS: 2197 molecules / 4.04 nm / 300 ps at seeds 11, 12, 13 and 8000
  / 6.21 nm / 200 ps; sampled every 4 fs (B286b's 20 fs was marginal above
  k ~ 5) on the centre-of-mass and atomic currents.
- THE RESULT: eta(k) rolls off as a GAUSSIAN (chi2 33 vs Lorentzian 192)
  with k_c = 7.89 / 7.89 / 7.94 / 7.93 -> 7.90 +- 0.05 nm^-1. c_0 (5.31):
  factor 1.49 - at k_c(c_0) the measured suppression is 14 pct where
  PAPER_2275 says 29. c_inf (2.45): factor 3.2, EXCLUDED. Shear waves begin
  at 1.0-1.4 nm^-1, four to five times below k_c(c_0).
- THE VERDICT (PAPER_2281 sec 4, as written before measuring): front 4
  CLOSED NEGATIVE ON THE COEFFICIENT - the 1/e wavenumber 13.5 lies far
  from both 9.05 and 4.19 - and POSITIVE ON EXISTENCE AND SHAPE. A failed
  prediction, recorded as one; what survives, stated separately.
- THE RULINGS (Daniel 2026-09-17, on the brief): Q-247 c_0 stands and the
  literal 1480 m/s is re-provenanced (PAPER_2267 L53; PAPER_2261 has no
  sound speed); Q-250 eta(k) on the centre-of-mass current is canonical;
  Q-246 the PAPER_1042 Gaussian tail is THE high-k form. Q-251 OPENED:
  k_c(measured)/k_c(c_0) = 1.489 +- 0.010 vs D_BSFG/D_phys = 1.5 -
  post-hoc, one liquid, one temperature; an observation, not a claim.

Not claimed: front 4 passed; the 8 nm / 1 ns plateau (not run - the
workspace suspends between turns; the 4 nm seeds and the 6 nm box were
chosen instead, on record); the value of 0.156 nm; anything about real
water beyond TIP4P/2005.

Gate 6,114/0. Dispatch 2,333 keys. Registry 7,137 rows. Acceptance
109/109 (simulator untouched).
