Paper 49D: An Alpha-Free Boundary–Interior Scale from Cuboctahedral Shell Refinement Analytic Model Limit, 7.6-Trillion-Event Hardening, and a Locked Electromagnetic–Nucleon Audit

Paper 49D develops an alpha-free dimensionless scale from a boundary-interior refinement model using the inherited shell counts 642 and 2562. Removing two declared closure positions from each shell gives the exact interior relation 2560 equals four times 640. A four-state boundary-interior orientation grammar cancels the mixed cross term under symmetric weighting. Six declared geometric residuals are then combined through an equal-weight root-mean-square operator and exponential retention map. The resulting analytic model limit is 0.007287307110890731. A production campaign tested the same frozen model through 10,000 balanced rounds, 760,000 tasks, and 7.6 trillion stochastic events. The locked holdout estimate was 0.007287301038212111, differing from the analytic limit by 0.833 parts per million. Development and holdout partitions, two independent random-number engines, temporal subdivisions, sensitivity variants, mechanism controls, and neighboring-shell controls satisfied their declared scientific contracts. Fifteen of sixteen total contracts passed; the sole failure was a runtime-duration requirement caused by reaching the fixed round cap. After the forward artifacts were locked, the analytic scale was compared with external electromagnetic and nucleon data. Its ratio to the fine-structure constant differs from the proton-to-neutron mass ratio by approximately 0.0668 parts per million. This establishes a precise alpha-free candidate relation, not a completed derivation of the fine-structure constant. The remaining requirements are a native derivation of the residual operator and an independent prediction of the proton-to-neutron mass ratio.