Two-channel scaling of a graph-spectral temperature proxy on a growing causal network

We measure a cone-correlator correlation-temperature proxy on canon, a stochastic grow ing network whose directed edges carry two tunable channels: local edges (density PRETURN) and wormholes (probability w). On the past cone of a vertex, a massless scalar field in the cone ground state defines, through the Peschel-kernel modular Hamiltonian K, a single scale w0 (T ≡ 1/w0): the Mp shell profile is single-scale with boundary/max = 1.00 in QC-passing runs up to 4338-node cones (N = 8, 12, 16); within a fixed regulator only ratios and exponents are physical. The pre-registered additive law fails on the full two-mass grid (R2 = 0.85–0.93; the unmodeled mass channel dominates); adding the mass channel plus one cross term is decisive by AIC (R2 = 0.969–0.976), still below the pre-registered 0.98 threshold—the full-grid threshold test remains primary, and fixed-mass fits are labeled di agnostics. At N=16: aP = 1.28 ± 0.15, bw = 0.46 ± 0.03, γT = 0.13 ± 0.02 stable across N=12–20. Sharing the same m → 0 divergence, the kernel behind K and w0 define a normalization window (drift ≤ 5.4% per mass decade, consistent with zero). Null controls bound interpretation: wormhole rewiring reproduces the surface within confidence intervals; scrambled cones fit equally smooth power laws with different exponents and no positive interference. Smooth scaling is generic; the discriminative content is the exponent family plus the interference sign.