v0.436.0 — THE REYNOLDS LADDER + SPINE AUDIT + THEOREM B SHIP

B278 (PAPER_2273): the cap across Re_lambda 433 -> 2,500, up to the
largest DNS in existence.

- A cap graded at one Reynolds number (B276) could be a cap on
  borrowed time. The ladder asks the question one snapshot cannot:
  does the sampled 1182-form statistic climb toward 17/20 with Re?
- Daniel supplied the sanctioned JHTDB testing token a third time.
  Same protocol - official REST service, deterministic bit-reproducible
  points (B276 anchors re-pulled bit-exact FIRST), 2 x 1,000 gradient
  tensors per rung, new seeds - on isotropic1024coarse (Re_lambda
  ~433), isotropic4096 (610.57), isotropic8192 (1200-1300, B276) and
  isotropic32768 (~2,500: 32,768^3, 3.5e13 grid points, the
  record-setting Frontier exascale run), plus a resolution check at
  fixed Re_lambda ~610 (4096^3 vs isotropic8192 snapshot 6 on 8192^3).
- THE CAP HOLDS AT EVERY RUNG: worst ratio anywhere 0.107323, eight
  times under 17/20, verdict by the tier-3 harness on 88 rows. The
  global statistic shows NO trend toward the cap - the highest rung
  carries the lowest mean. The resolution pair agrees within seed
  scatter. Positive-stretching fraction in the textbook 0.75-0.78 band
  at every rung.
- Rule 7, the headline: the 100-point sub-batch envelope drifts
  monotonically upward with Re (0.068 -> 0.107) - the intermittency
  direction, exactly where a violation would live. FLAGGED for the
  full-token extreme-event scan, not extrapolated away. The ladder
  answers the Re-trend question, NOT the kill test, which stays OPEN.
  isotropic4096's transient backend faults are disclosed; every
  recorded number came from a 200 response.
- Daniel ledger unchanged: JHTDB full token / Kerr letter = the kill
  test + the pair-cap discrimination (+ now: does the envelope
  saturate?); THz bench; GFZ deep-sonic; U_i spec; SEG-Y; quartz
  bench; tester answers.

B279 (same band) - THE SPINE AUDIT (PAPER_2274): the proof-gap ledger
opened at B269 still carried one unnamed theory row, ns_functional_
spine. Daniel: "find the missing pieces" in the predecessor repo. It
was the predecessor S300 "UQFF-Leray" argument, and it is audited here
with file:line quotes: the budget and the cap stand (derived since
B271), the Sobolev step is FALSE - dimensionally and because H^1(R^3)
is not in L^inf (a divergence-free family with finite E, finite E_2
and unbounded sup is exhibited and witnessed live) - Young yields only
Leray's small-data condition, and a constant 17/20 leaves the BKM
criterion untouched. The predecessor's own audit had said ASSERTION_
ONLY. Theorem A never needed the step; the domain ruling assigns the
regime that does to mathematics; the row closes BY THAT RULING with
this paper as the record. Registry open theory rows: ZERO. The Clay
statement is not claimed.

B280 (same band) - THEOREM B (PAPER_2275). Daniel: "I have a feeling
the solution is all three. Let's solve for all three." The corpus
carries the phonon sector in three forms - the LINE (Gaussian /
Lorentzian at f_c = 1.25 THz, Q = 25/2), the THRESHOLD (H_SCm: dynamics
only below the carrier) and the Gaussian TAIL in mode number q^{n^2},
q = exp(-beta_i [SSq]) = 0.7092 (PAPER_1042) - and composed they are
ONE Fourier multiplier: unity below k_c, a shoulder at k_c, Gaussian
above. In x that is a Gaussian mollifier of width eps = sqrt(2 beta_i
[SSq])/k_c = 0.156 nm for water - no free parameter, one flagged
identification (n = k/k_c). Put on the transport, where the corpus
puts it, the CONTINUUM fluid - all modes, nothing truncated, no ruling
needed - is Leray's 1934 regularised system: energy conserved; the
smoothed field's gradient bounded by the energy, ||grad u_eps||_inf <=
C eps^-5/2 ||u_0||_2 (the step S300 needed and could not have,
legitimate because the bounded field is the SMOOTHED one; C^2 =
3/(16 pi^3/2) EXACT); Gronwall closes H^1; the solution is UNIQUE,
GLOBAL and C^inf. No cap, no envelope, no Sobolev embedding, no BKM.
Theorem A is the sharp-filter limit; the constant diverging as eps ->
0 is the domain ruling made quantitative. The whole-corpus sweep
found the same threshold twice - the PAPER_106 power-law forms close
the same track iff beta > 5/2 (mollifier) and a >= 5/2 (Lions) - so
Track 3 is CLOSED on the Gaussian form and CONDITIONAL on PAPER_106,
pending Daniel's exponent ruling. Front 4 gains a shape (Gaussian
roll-off, 1/e at 1.71 k_c). The Clay statement is not claimed.

Gate 6,105/0. Dispatch 2,326 keys. Registry 7,119 rows. Acceptance
109/109 (simulator v1.90.0 untouched).
