Finite-Size Indeterminacy in a Coherence-Landscape Scrambling Probe: Exact Diagonalization Evidence from MERIDIANO-v1

TECHNICAL RECORD — reproducible experiment. This paper does NOT test whether the RCD coherence field C is fundamental or physically real. It reports a finite-size scrambling diagnostic applied to a correlated-disorder coupling landscape generated within the RCD programme, and documents why, at exact-diagonalization sizes, that diagnostic returns no verdict. This bound holds in both directions: even a future PASS would be a result about scrambling in correlated disorder, not evidence that C is a physical field. The programme-level interpretation of this episode is given in the companion Reframing paper; here we provide only the technical record and its reproduction path. MERIDIANO-v1 asks whether the specific spatial ordering of an RCD-generated correlated coupling landscape leaves an irreducible imprint on operator spreading in a non-integrable mixed-field Ising chain, beyond what the distribution of coupling values alone produces. Using exact diagonalization at L = 12 with all parameters and death conditions pre-registered, we find: the chaoticity gate passes (level-spacing ratio <r> = 0.5348, GOE, after correcting an unresolved spatial-reflection symmetry); the raw operator-front width measures a broadly-universal diffusive broadening not specific to the landscape, and is reframed to the exact deviation of the front from a calibrated local-velocity (eikonal) prediction via a double difference that cancels a convention offset; at a single interface the deviation is localized at the wall (peak |SIG_xf| = 0.362 sites); but across a six-landscape scatter the signed deviation is sign-unstable (CV = 2.64), with the positive case border-contaminated. The governing summary: there is an interface response at L = 12; there is not yet a reproducible interface law. The verdict is INDETERMINATE by finite size — at L = 12 one cannot distinguish absence of a reproducible local law from an instrument too small to resolve one. An initial NO-GO verdict, co-signed without friction by four adversarial models, was corrected to INDETERMINATE on the ground that the sign instability is exactly the signature of finite-size interference; the correction is recorded as a documentary fact. We specify the resolution path — a system-size sweep with numerically pre-declared persistence thresholds (named but deliberately not fixed here, since the sweep is pre-registered on its own sheet), made feasible by dynamical-typicality estimation of infinite-temperature OTOCs validated against the exact L = 12 datasets — with three frozen outcomes (PASS / physical NO-GO / method-INDETERMINATE). Rescue-by-MPS of a fragile signal is prohibited; sizing the instrument is a distinct, legitimate operation. No physical status is claimed for C. In its best case this line yields a self-standing result in the physics of scrambling in correlated disorder.