Within the Radial Coherential Dynamics (RCD) program, we test whether a phenomenological macroscopic bridge between the dynamics of a coherential scalar field C (x, t) — governed by an Allen-Cahn drift with multiplicative Wright-Fisher noise — and the experimentally measured Margenau-Hill quasiprobability negativity on IBM superconducting hardware can be constructed using differential observables based on spatial gradients of C. Three independent observables in the gradient family are evaluated: an instantaneous integral SₙablaC = ∫|∇C|² dx, a temporal-integral form Wᵢnt accumulating the instantaneous form across the simulation window, and a pointwise maximum SₙablaₘaxC = maxₓ |∂ₓ C|. All three observables are normalized as differential ratios between coherent and control preparations. Synthetic tests with three controlled decay regimes (exponential, linear, threshold) show Pearson correlation between exponential and linear regime curves of +0. 9995, +0. 9999, and +0. 9995 respectively for the three observables — virtually identical kinetic responses for functionally distinct decay forms. We conclude that gradient-based phenomenological observables with differential cociente normalization are structurally kinetic-blind: they detect monotonic decrease of structure but cannot distinguish the form of the decay. High correlation with an experimental survival curve is not sufficient evidence of kinetic agreement when both curves are monotone. This closes the gradient family as a candidate for the macroscopic-microscopic phenomenological bridge in this program, while leaving open non-gradient observables and microscopic-dictionary derivations as future routes. The result extends the previous registered negative result of v1. 0 (Zenodo 10. 5281/zenodo. 20073982) from a specific observable failure to a structural limitation of an entire observable family, with full ex-ante kill-switch protocol and synthetic adversarial testing.
Arturo Cerezo (Fri,) studied this question.