Uses the saved design, not hard-coded EM names, HCR counts, or iteration IDs. Replays each procedure's own HCR and failure policy against its catch history. Validation does not certify scientific adequacy, model convergence, or risk acceptability. Diagnostic flags and missing EM estimates remain visible.
Arguments
- result
A completed
run_mse()result.- tolerance
Absolute numerical tolerance for replay and common initial state.
- stop_on_failure
Stop if any computational check fails.
Value
List with passed, named checks, target_audit, and
exploitation_diagnostics from summarise_mse_diagnostics(). Guard events
do not themselves fail validation; inconsistent saved history flags do.