SPAR is not a physics-only framework.
Physics is where the framework first proved itself.
That matters because mathematical and physics-grade systems make claim drift visible in a particularly hard form:
- analytical contracts matter
- approximation regimes matter
- maturity state changes what the result is allowed to claim
- reproducibility alone is not enough
In Flamehaven-TOE, the review problem was not simply whether the engine ran. The harder problem was whether the output still deserved the claim attached to it.
That produced concrete failure modes:
- a path can return acceptable-looking values while remaining epistemically empty
- a formerly heuristic path can become genuine while the registry remains stale
- a governance score can look smooth before its underlying formula deserves the confidence attached to it
The clearest example is SIDRCE Omega.
An earlier version used a large arbitrary multiplicative constant on a raw residual. The score behaved smoothly, but the formula was stronger in presentation than in justification.
It was later replaced with a chi-squared Gaussian construction. That changed the epistemic status of the score because the reported value became reversibly connected to the underlying residual:
||beta|| = tol * sqrt(-2 * ln(score))
That is exactly the kind of shift SPAR is meant to notice and classify.
The other proof point is the registry itself.
SPAR keeps maturity and gap states as machine-readable review objects rather than prose-only caveats. That allows the runtime to emit:
- model registry snapshots
- gap registry snapshots
- maturity state surfaces that travel with the result
This is the bridge between concept and operational review.