Empirical Regime Taxonomy in Boundary-Constrained Generative Systems

We analyze 570 unique shapes generated by a discrete constrained trajectory system in which each step irreversibly consumes combinatorial resources. Seven statistically stable dynamic regimes emerge from unsupervised clustering. The central finding is threefold: geometrically identical configurations can belong to dynamically distinct regimes, all trajectories terminate in one of exactly three outcome classes (Closure, Sustained Cycle, Progressive Infeasibility), and the terminal outcome is anticipatable from early dynamic signals. Failure in constrained systems is finite, structured, and anticipatable. Part of the sub-limit dynamics research series.