Localized Cascade Risk at the 16 September 2026 Model-D Phase Boundary: A Pre-Registered Fast-Slow Stabilization and Falsification Protocol

This pre-registration fixes a limited-disclosure Model-D hypothesis before a model-defined 16 September 2026 phase boundary. It tests whether localized thermosphere/ionosphere residual structure can be materially stronger than globally smoothed representations suggest, and whether a predeclared fast/slow representation improves holdout performance over a single-timescale baseline. Established space-weather literature supports strong spatial and temporal heterogeneity in thermospheric density and nontrivial storm-time model uncertainty; it does not establish the Model-D date, the fixed internal anchor mu=0.218, or a deterministic cascade. The release defines explicit falsification criteria and a model-independent emergency fallback while withholding the full live-state resolver, private joining/synchronization logic, operational thresholds and any actuation code.