It ought, one imagines, be possible to construct a model of how one believes the Flat Earth to work. Certainly one can do that for a spherical earth, and get a very good fit with observation regarding sunrise, trajectory and sunset. That way you don't have to have any fancy physics, except with a small globe all the water would fall off.
So, take your pizza base and shine your torch at it. At equinox, half the earth is in a shadow which bisects the north pole. You can ring people up and check. At the equator, the sun is overhead at noon, so logically your torch needs to be over that line on your plane (I'm assuming the North-Pole-centered model here). Is there anywhere you can place your torch such that half the disk is in shadow (ie an ant on the pizza could not see it), and it is overhead at the equatorial line somewhere? One could introduce a system of baffles, but then the ant would see the baffles, rather than what is behind them, and the sun would be above the horizon when it went behind them, which is not how we experience sunset or starlight.
An ant on the equatorial line on a spherical earth would see the sun rise straight up, pass directly overhead, then sink directly down the opposite horizon. On a flat earth it would see it moving one way in the early morning, then hang there for a while before accelerating in the opposite direction to occupy a position overhead. After noon, the morning shenanigans would be reversed. The sun would not rise and set at opposite horizons.
Matters become even trickier to engineer at the southern hemisphere's summer solstice. One would need to allow the entire Antartctic rim to be bathed in light, but still have a special cowl that admitted starlight but not sunlight (or had the stars painted on) over half the surface. Or bendy light, whose existence certainly makes the Round Earth Conspiracy easier to sustain.
Or, I dunno, maybe spherical models work because the earth really is spherical?