Cross-platform replay of the frozen assessor returned INVALID_SET because
recomputed eps(128) differed from the Horsea receipt by one ULP. Compare
only that derived float within 4 ULPs (close_ulps); hashes, case identity,
other metadata and scientific thresholds stay exact. Tests accept the
observed one/two-ULP differences and reject 1e-12 relative, 2x, string,
NaN and missing values.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- run_case: --diagnostic scaled_shear_n256 (requires --flow shear_wave;
U256/U512/constant R256/projected R256, Re100, no early trace); N512
only via one of the two scaled diagnostics; scaled_n256 stays TG-only.
- assess_shear_wave: scaled=True contract reusing 92b8afe scaled
case/build/workload checks, shear diagnostic identity; signed rate bias,
refined excess over each control; hash-pinned de5b0cc N128 reference and
descriptive N128->N256 signed trend; CLI --diagnostic/--reference-files.
- tests: scaled contract, set mixing, trend/reference pin, runner options.
- shear_wave.py and oracle.py unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>