The rub, as I mentioned before, is that no matter the angle, the buildings follow the classical laws of perspective, not Rowbotham's (unless the horizon begins to block the lower portion of the building, at which point it appears to follow Rowbotham's rules as the lower portion of the building goes out of sight long before the rest of the building becomes to small to see).
Rowbotham's perspective doesn't work except that it does. I immediately cede my understanding to your well considered opinion.
Are you incapable of understanding what you read?
Objects are obscured by the horizon as they pass beyond a certain distance (dependent on certain local variables, such as terrain, the height of the object, and the height of the observer). Such objects cannot be restored by looking through a telescope (I know, I've tried it), but can be brought back into sight by increasing the observer's altitude (again, been there done that, and may I say, it's pretty cool watching things 'roll' into view as you climb higher and higher!). Invariably, the object appears smaller as it gets further away,
in accordance with the classical rules of perspective, and only ever disappears from view when it is so far away that it becomes indiscernible (the smaller features - such as masts on sailing ships - having long since been lost to sight), or when it is obstructed by the horizon.
EDIT: >> A thought that occurred to me after posting this: For Rowbotham's perspective to work as he says it does, where objects further from the eye line shrink to invisibility less rapidly than those closer to the eye line, objects such as buildings should actually appear to taper toward their bases as they get further away, with the base rapidly tapering into invisibility at the horizon line. This is definitely not what is seen. <<
Essentially, Rowbotham's perspective is a failed attempt to provide an alternative explanation of why objects disappear as they pass over the horizon. It defies logic and contradicts itself, as Alex pointed out in the OP.
I strongly encourage experimentation.
Then I suggest you do some of your own.