M03 independent computational review
===================================

Independent means a separate computational implementation within this project,
not an outside laboratory, external replication, or measured specimen test.

The prospective review plan fixes two reference solves before outcomes: meshes
n10 and n12 at index80 of the original161-point logarithmic frequency grid,
63,245.55320336759Hz. The main author sweep must finish before these two guarded,
sequential workers execute. No result-selected peak or replacement point is used.

Three coverage levels are kept separate:

1. All322 scalar/report records are checked for exact identities, file hashes,
   frequency assignments, statuses, reported validity gates, scalar energy/power
   identities, parent replay, curve exports, normalization and summary agreement.
   This does not independently reconstruct the316 full fields that were not saved.

2. Six saved displacement fields (indexes0/80/160 on both meshes) are independently
   reconstructed. The reviewer builds Cartesian shape gradients, tensor strain,
   tensor stress, volume/time RMS von Mises stress, consistent-element mass,
   elastic/kinetic energy, power and element-wise internal force. This provides
   a second dynamic-residual calculation without importing the author's FEM code.

3. Only the two prospectively fixed midpoint points are independently resolved.
   The separate Cartesian tensor element implementation builds K/M/F and calls
   the installed SciPy sparse direct solver. A shared external linear-solver
   library is disclosed: this is separate assembly/field-recovery code, not a
   claim of independent solver-library diversity. Full displacement differences
   are measured in the consistent-mass norm, and scalar compliance/stress are
   compared. The reviewer tolerance1e-8 is a consistency tolerance, separate
   from the study's unchanged5% accuracy screens.

The main model remains a homogeneous isotropic20mm steel cube with its bottom
fully fixed and a1N peak uniform top load. Rayleigh damping is an illustrative
assumption, not measured loss. Coordinates and displacements are metres, elastic
moduli and stress are pascals, mass is kilograms, frequency is hertz, stored energy
is joules and power is watts. The harmonic convention is exp(+i*omega*t).

Global RMS stress averages over both volume and time. It is not a simultaneous
maximum stress or a fracture test. The denominator for the original M02 modal
mesh metric is based on Hm12; the M03 decomposition uses a common Hd12 scale.
Their percentages are not interchangeable. Complex cancellation in
M=D+E10-E12 is an algebraic identity, not a partition of energy or causal shares.

The direct n10/n12 difference is a difference between two discrete meshes,
not the true continuum error. Even all161 sampled frequencies do not bound
continuous-frequency extrema between samples. Scalar compliance agreement does
not prove convergence of every displacement or stress component. Numerical
validity can pass while discretization accuracy fails.

The resource guard enforces a configured Windows Job aggregate worker commit
limit. Resident and combined coordinator/worker usage are sampled. The native
peak-counter anomaly from the earlier allocation-denial diagnostic remains
unexplained and is not treated as proof of admitted memory. The900second main
budget stops monitored work; cleanup/final serialization can overrun it, and the
finalization receipt records that explicitly. Independent workers have their
own predeclared120second total and55second per-worker limits and do not overlap
the main sweep.

No hardware trial, fracture capability, acoustic delivery device, patentability,
manufacturing qualification, or general material/size transfer is established.
