Statistical Hybrid Riemannian-Ensemble Dimensional Integration (SHREDI) Reveals Metabolic Arrest in Financial Manifolds

Financial stability frameworks assume that crises resemble ecological tipping points, where rising autocorrelation signals impending collapse. This paradigm fails for infrastructure-constrained markets, which exhibit a synchronisation paradox: behavioural convergence without volatility amplification. This study introduces the Statistical Hybrid Riemannian-Ensemble Dimensional Integration (SHREDI) framework to diagnose this condition, termed metabolic arrest. Integrating Random Forest classification with topological data analysis and Riemannian geometry, SHREDI analyses South African financial markets during sustained electricity load-shedding from January 2015 to December 2025 (N = 2,731 trading days). The study establishes that Stage 4 load-shedding, corresponding to a 4,000 MW generation deficit, constitutes a deterministic threshold: effective dimensionality collapses, autocorrelation declines rather than rises, and entropy bifurcates discontinuously at a magnitude consistent with the universal range documented across jamming transitions in physical systems. Machine learning achieves classification accuracy unprecedented in financial econometrics, and Fisher Information Gain confirms the transition from probabilistic regime overlap to deterministic constraint. These results invalidate ecological early-warning frameworks and establish that infrastructure-driven fragility belongs to the jammed universality class, requiring macroprudential surveillance to shift from monitoring weakening restoring forces to monitoring geometric constraints on market mobility.