Qualify transport with reused exact certificates

The first component-transport study completed with 20/24 passing controls; four broad globe bounds reached their unchanged 60-second limits. Preserve that freeze, all charged work and its failed qualification.

The new candidate changes bookkeeping only. Reuse immutable, already validated directional views and keep a bounded cache of repeated exact norm calculations. Do not change the rational arithmetic, rounding allowances, model equations, calibration domains, native centers or acceptance thresholds. Local equality tests compare every returned norm and component bound bit for bit with the original component implementation; training-region tests retain all 51 columns.

Repeat the unchanged 24-case native control matrix for this separately frozen candidate: six derivative, six packet-boundary, six broad-domain and six stationary cases, covering all three physical candidates and both processing assumptions. Keep 1 micrometre center/stationary limits, the original .002 finite-difference step and 20 micrometre derivative limit, and the original 10 micrometre nonlinear-containment margin.

Finite allowance: 24 identities, 78 native evaluations, 19,632 native packet visits, 348 bounds/source certificates, 15,072 bound packet-steps, no retries or extensions. Use two workers with two numerical threads each to reduce competition within each bound's unchanged wall-time limit. The overall ceiling remains 600 seconds; each bound remains limited to 60 seconds. Freeze sources, inputs, compiler and environment before launch. Reserve work before execution.

No original aircraft reads, ESA confirmation data, observed fitting or optimizer starts. The norm-constrained body-source interval is still a separate integration task. This study does not authorize a complete prediction rerun or aircraft fits.