Common-clock force integration qualification
Qualify the exact integral and clocked-matrix components before native use. Use six unchanged generated configurations and four cases each: a 16-packet short clock domain, a 400-packet full clock domain, a 1,600-packet full clock domain, and a 1,600-packet fixed clock. Short-domain sensitivity is a software control, not a change to any aircraft calibration limit.
Each of 24 identities uses 12 predetermined/bounded parameter probes. Compare the integral centres with an independent backward suffix-weight calculation. Check the exact rational G/H remainder inequalities on all three axes at each probe. Use an independent Rodrigues/reciprocal-gain matrix for the weighted prediction. Retain all 27 physical derivative columns and the 51-column design.
Finite-step derivative checks use step 2e-6 and tolerance 3e-8 in the generated output units. They are software checks, not a claim of a unique differential at clipping boundaries. The declared clock linearization is protected by its exact remainder. Matrix-weighted containment allows 3e-12 for the independent floating oracle; it does not establish full floating-point interval certification.
Freeze 24 starts, 288 probes, 1,296 derivative evaluations, 1,728 exact G/H inequalities, at most 336 independent history queries and 325,440 combined packet visits. Cache a history only when its exact query time is identical; 26 physical directions leave the clock unchanged. Reserve before work. Four workers use one numerical thread each; allow 180 cooperative seconds, 30 per integral calculation, zero native calls and zero native bounds. No original aircraft reads, observed fits, retries, extensions or process signals.
Keep source/input/environment hashes and every failure. Component success does not release native propagation, the typed prediction interface or aircraft fits.