Conditional continuum analogue only; no actual native trajectory error or continuum-limit proof. Reference remains17/19 qualified.

One-shot run: python thin_film/scripts/closure_comparison/run.py --prepare, then run.py, with one BLAS thread.
Presentation: python thin_film/scripts/closure_comparison/report.py.
Original reference files and all9 modal state hashes verified, no reference/native evolution replay.
Metric: equal-weight sample relative complex L2 over the original ten times, not time-integrated L2 or original C1 acceptance norm.
No model-agreement pass threshold. Original actual C1 scope/gates remain.
Git may normalize processed text line endings; on-machine hashes refer to original local text.