What would a flat-Earth Moon do?

Choose your Moon before the demonstration.

Let an advocate set out the strongest version in their own terms. The linked map proposal is one example, not a spokesperson for every flat-Earth claim. Once a model is chosen, keep it chosen. A nearby lamp and an observer-specific image don't owe the same explanations.

Now make a small prediction sheet: direction from two locations at one time, illuminated fraction, orientation of the lit boundary. Those questions connect the argument to a shared sky. A model can be imaginative and still be asked to specify a result before somebody looks up.

Don't confuse refusing authority with refusing measurement. The first can motivate an excellent investigation. The second turns every possible observation into another reason to believe. A theory becomes more interesting when it supplies a result that would make its own author reconsider.

Pick a model, not a mood.

The linked geocentric proposal is one claimant's account. We do not assume it represents every person using the label flat Earth. The useful first step is to specify the lunar distance, path, illumination and observer geometry.

A model should give predictions before seeing the observations. Otherwise a perspective adjustment can become a rescue device rather than an explanation.

Make the constraints cooperate.

The Naval Observatory describes phase geometry; NASA's eclipse guide describes the alignments producing eclipses. A lunar model needs to account for both patterns, not answer each with a separate incompatible diagram.

Our proposed stress test adds two simultaneous observers. The same Moon must produce their different viewing directions without changing the model after each photograph arrives.