Results/source: https://github.com/wang-mengdi/variational-films/tree/623be3ae3ebe2709ea621a1b084a5b850b3a8fdc

# Actual uniform5 startup provenance

Native plan/driver:4c05fd9. Independent audit wrappers:c2b3ce1.
Sources: thin_film/scripts/global_five/. Original native subset hashes and
bench/unit binary hashes in protocol.json; original24test logs verified by
name and exact counts(observer6,preparation4,curved14), no unit rerun.
No C++ edit or native build in this campaign. Production binary remains
76e49ed253e2341df37f93d02803272d020b8500ed9265da6f7d0ec124b6c5c3.

Both native cases terminal before auditing. Existing independent SI/GEOS
scalar+mechanical algorithms reused WITHOUT changed formulas or tolerances,
on exact immutable copies outside native-run directories. Each case verified
seal records raw/core-source/processed-output hashes. Wrapper source inventory
was recorded after execution(auditor_source_verification.json), explicitly.
Rendered/processed text can undergo Git newline normalization; exact original
NATIVE data are preserved inside bounded ZIPs and restored via SHA256.

Archive namespace native/ restores to results/local-runs/global_five_20261113/.
Initial10um visual reference from ../global-space/,guard_dt00625; source SHA
in layout_sources.json verified, no reference trajectory rerun. No computed
10us uniform5 trajectory exists in this batch. Both one-step horizons differ.
All native/audit stdout logs retained; no timeout or native failure occurred.

Native total209.313s; independent audits6.470s. Per native hard cap600s,total1200s;
per audit300s,total600s. No>2h job launched, no long-job monitor needed for
these capped starts. Future>2h launches require a half-hour monitor first.
Cost estimates from initial step are engineering projections, not controlled
performance comparisons or predictions of nonlinear future iteration counts.

Full C1 scope unchanged; single-step numerical closure does not establish
physical normal flux/traction, actual10/50/200us convergence or experiments.
Energy Hessian dense-outer cost is a hypothesis from code inspection only.
Future work keeps generic solver/backend ownership untouched; DRIFT-owned
energy assembly alternatives require independent equivalence controls.