The solution to being unable to distinguish between the two rival interpretations of the Eratosthenes experiment is that if you do the measurement from more than two locations, you get different answers in one model, and the same answer in the other model. This argues for either one model to be wrong, or for Magic Perspective to come to the rescue of the failing model, and sheer coincidence to allow the other model to be wrong though it predicts consistent answers.
So a single Eratosthenes experiment can't distinguish between the two situations in your model, the iterated Eratosthenese experiment can, by getting conflicting values for the distance to the Sun.
The sun moves 15 degrees every hour. At the equator, 15 degrees of longitude is about 1000 miles. if you did an east-west Eratosthenes experiment where you measure the angle to the sun at a given time over 4 different timezones, you would find the following distances to the sun:
1000 miles / tangent(15 degrees) = 3700 miles
2000 miles / tangent(30 degrees) = 3500 miles
3000 miles / tangent(45 degrees) = 3000 miles
4000 miles / tangent(60 degrees) = 2300 miles
Of course, this would resolve to the sunset problem - at 90 degrees, the sun is touching the earth. The solution to the sunset problem is "perception" so I guess they would give you the same answer to this. And, "perspective" changing in ways that nobody has a model for means that the answer to how far the sun is from the earth is "nobody has any idea."
This might bother someone who thought they knew how far it was from, say, New York to Paris, but if you disclaim knowledge of such things then why would you be worried about the sun possibly being closer to the earth than San Francisco is to New York?