Directional Access in a Fixed Potential Landscape: A Two-Operator Model of Diffusion Geometry, Transient Reactivity, and First-Passage Accessibility

Selective access may depend on which dynamical directions are opened, not only on how strongly activity is amplified or suppressed. We test a two-operator stochastic model that separates deterministic signal propagation Asig from stochastic access Anoise while holding a two-dimensional double-well potential fixed. Trace-matched anisotropic diffusion rotated short-horizon dispersion (mean aligned-versus-opposite target-alignment margin 0.262; maximum rotation error 0.0186 rad). Spectrum-matched CT versus rotated CR is the principal signal contrast: CR produced Euclidean stable reactivity (α = −0.1647, ω2 = 0.3317) while CT remained contractive (α = ω2 = −0.1). Euclidean finite-horizon gain reached G2(0.5) = 1.0764; the same growth is absent in the Asig−1 metric, where GM(τ) = exp(ατ) < 1. Relative signal–noise orientation increased finite-horizon covariance along the full tested curve, reaching Δint = 0.1076 at the upper boundary τ = 1. TA-5 v1.2 recovered a reflected-domain first-passage orientation effect (ΔqPDE(+) = 0.0621; ΔqMC(+) = 0.0555) against a near-null identity baseline. This is a bounded synthetic qualification under a frozen Euclidean state-space contract, not a biological thalamus, TRN, memory, consciousness, or NDT-macro claim. TA-7A-ND remains development-only; TA-7B remains closed.