The reference must preserve signed energy transport
The richer local diagnostic saved its complete collection before offline tracking failed. Repeated input filenames collided in a tracker that copies by basename. The original scheduler failure stays recorded. An archive with relative case paths now retains the original files, their provenance and checksum; recovery required no additional solver calls.
| evidence | conclusion |
|---|---|
| direct signed collection and saved-output audit | 648 integrity checks pass on the complete 15-state collection; all affected signed states agree between backends, but four control states fail the existing parity criterion; original terminal tracking failure is preserved and archive recovery calls no solverC-199 |
| repeatability tooling | 31 synthetic diagnostic tests pass in each mode, including archive identity and repeatability completeness/discrepancy checks; 33 unchanged focused records are retained, giving 64 focused recorded passes per modeC-200 |
| coupled reference acceptance | unresolved; the original uncovered face remains |
The collection confirms signed energy flux where the previous coefficient inversion had only inferred it. It also reveals backend discrepancies in preceding controls. Agreement at the affected states does not excuse disagreement elsewhere, and a successful spectral integration is not an independent physics validation.
The owner delegated a literature-based review of what the reference should be. Primary TGLF documentation defines species energy fluxes and separate mode outputs. Published validation studies compare matched inputs, spectra, fluxes and coupled profiles. Literature also contains mechanisms for inward energy transport, but their existence does not identify the mechanism in these particular inputs.
The approved method preserves signed, separate-species TGLF transport as the target reference and requires numerical reproducibility before a coupled comparison. The current guarded trajectory remains a regression baseline. Clipping the signed output would define a different modelling baseline; it cannot retroactively supply missing reference coverage. A conservative stable energy closure needs its own concrete implementation review. The served surrogate contract and current guard stay unchanged. M4 and its temperature gate remain open.
What’s next. Run a frozen bounded repeatability check, keeping each fresh backend evaluation and every discrepancy. Then review numerical convergence and the energy closure. No remote artifact synchronization has occurred.
Technical details → Verification › signed reference literature review
Decisions → OQ-33 · OQ-34