E01 corrected-volume independent review

The independent audit agrees with the revised saved evidence, INCLUDING the
reported failed production gates. Agreement is not a claim that every run
passes. The current composite contains one failed acoustic nonnegative-ledger
gate and two failed impact boundary runs, as well as fourteen additional strict
reviewer monotonicity findings.

Why a new version was necessary
The initial acoustic enclosure had tangent mounting bosses: a kernel-valid
shape still contained five disconnected solids. Correcting boss radius from
6 to 7 mm creates nominal wall overlap and changes the modeled air volume from
0.706091410953079 L to 0.7039780004475063 L, a reduction of about 0.2993%.
The original physics files remain historical evidence for the old volume.

The revised composite recomputes 486 ND64 algebraic cases and twelve ND64 ODE
runs. It inherits 486 ND91 algebraic cases, twelve ND91 ODE runs and sixteen
impact runs without rerunning them. There are forty integration records in the
current composite, of which twelve are newly calculated and twenty-eight are
inherited. Across the original matrix and the addendum there were fifty-two
matrix integrations. No hardware trials were performed.

Independent checks
verify_saved_e01_revision.py imports no production numerical engine. It checks
all 972 composite cases against a direct 3x3 complex solution and the actual
browser module, reconstructs all 24 acoustic and 16 impact retained traces,
and checks the exact inherited result/NPZ hashes, unchanged source hashes,
coarse/fine comparison fields, declared counts and CAD-volume input hash.

The final detailed audit has zero calculation/provenance mismatches in those
checks. Its status is AUDIT_MATCHES_REPORTED_PRODUCTION_FAILURES. The v1 review
is preserved but its headline wording was too easy to misread as every original
gate passing. The v2 review explicitly separates agreement with saved evidence
from the outcomes of the production tests.

Failed acoustic production gate
A_ND64-4_CAD_nominal_low20_coarse_volume_v2 has a minimum returned-supply-energy
sample of -3.1023928854649686e-12 J at t=0.0125 s. The frozen production criterion
requires every split-energy ledger sample to be at least -1e-12 J, so this
specific run FAILS that criterion. Its fine-resolution counterpart is above
the threshold. The composite acoustic_numerical_checks_passed flag remains
false. The run otherwise passes the declared response, phase and energy-balance
checks. A tiny magnitude does not erase the failed gate or justify changing
its threshold after seeing the result.

This negative initial returned-energy value is distinct from the additional
strict reviewer monotonicity criterion. Fourteen composite acoustic traces have
retained cumulative-energy decreases larger than 1e-10 J. The largest remains
1.34244e-8 J in the inherited ND91 one-litre, 1000 Hz coarse case. No retained
values were clipped, replaced or monotonized. Dense-output interpolation near
nonsmooth sign changes is a plausible mechanism, not a physical measurement.

Failed impact production gate
The unchanged deliberately soft capture case I04 exceeds the permitted lower
travel boundary at both resolutions. Its two failed integrations remain in the
composite, and no hard-stop load or successful capture is inferred afterward.

Provenance-label issue retained
Twelve new per-ODE ND64 detail files copied the configuration label
new_corrected_volume_algebra into their evidence_origin field. Their summary
records correctly state new_corrected_volume_integration; actual state traces,
solver metadata and hashes exist. The immutable detail files are retained with
this documented label error. Summary-record provenance identifies the run type.

The corrected CAD volume remains a Boolean proxy, not a measured acoustic
volume. Terminal geometry, gasket compression, wiring, leakage, actual driver
variation and enclosure compliance still require bench characterization. No
fracture, acoustic focus, hardware efficiency or physical capture performance
is established by this revision.
