v1.1.0

Adds six pre-registered experiments extending the method to external solvers and measuring its sensitivity boundaries. Each fixes its protocol and acceptance thresholds before any outcome is computed, and writes a machine-readable table or JSON report.

experiment	result
library_multires_audit	py-pde 0.56.0 Euler vs Runge-Kutta from output alone; grouped leave-one-out accuracy 1.000, recovered median rates 1.076 and 1.991
thermal_mesh_sensitivity	detection limits stable across a fourfold change in working mesh; measured false-alarm rates 0.042 to 0.064
battery_coupling_sensitivity	conjugate-coupling ladder, k = 5 to 80 W/(m K)
structured_noise_extended	structured-noise limits on a widened detuning range
openfoam_3d_audit	OpenFOAM v2506 scalarTransportFoam, three-dimensional transient, 60 runs, grouped accuracy 0.95
openfoam_turbulent_audit	OpenFOAM simpleFoam k-omega SST, 60 runs, grouped accuracy 1.000, minimum eddy/molecular viscosity 112

The two OpenFOAM audits run in a container pinned by image digest, refuse to start if the image identity or reported solver version does not match, and parse the discretization scheme back out of each generated case rather than trusting the value they wrote.

Solver-field caches under results/cache are regenerable with --cache and are not tracked.

v1.0.0

Initial submission release.