Probabilistic Drift Geometry Under Declared Constraint

This study introduces repeated-run probabilistic drift analysis under invariant artifact and contract conditions across three configuration states within a single kernel lineage: instrumented restraint (K0 + K0G-M), measurement-only instrumentation (K0G-M), and instrumental governance (K0G). Fifteen independent runs were executed and aggregated within an Operational Boundary Metrics (OBM) measurement surface. Results demonstrate that probabilistic drift geometry is configuration-sensitive. Instrumented restraint produces the smallest dispersion envelope. Measurement-only states expose maximal structured variance. Governance produces state-dependent compression relative to elevated dispersion but does not compress below the instrumented restraint baseline (K0 + K0G-M). These findings are descriptive and confined to the declared artifact and measurement grammar. The paper formalizes probabilistic drift geometry as a structured object consisting of dispersion magnitude, layer concentration, regime stability, deviation-set coherence, and interaction curvature across configurations. The contribution is methodological rather than prescriptive: architectural configuration measurably alters the probabilistic envelope of inference under invariant conditions. Transparency of boundary logic, instrumentation state, and governance activation becomes empirically relevant under configuration-sensitive drift. No normative ranking or task-class mapping is proposed.