An interesting Q about rotating objects

So I was trying to create a "rotating platform" functionality for my active platforms in my game. Now with normal platforms, what I do is I record the "Difference" in X & Y positions of a platform each frame, and move the player by that much (if the player was standing on it).

So thats pretty good for giving you free movement and proper platforming on moving objects. BUT, I want to go beyond that. I was wondering about creating platforms that could rotate underneath you, and would move you along the rotation axis.

But another thing about my game is that it uses strictly a-posterori movement, meaning you can never be overlapping an obstactle, and if you are overlapping one in one frame of gameplay, it moves you to your previous position (and if that position is now an obstactle, stuff.... messes up).

So I'm wondering about the feasibility of calculating the difference in a rotation of an active object. For example, I could find the angle/distance to the hot spot of the platform from the player, add the same degrees, and convert back from polar to cartesian coordinates and reposition the player.

But the problem is thats only an *approximation*. If you move just 1 pixel into the platform, it would detect it and move you to your previous spot. But as that is now occupied by the rotated section of a platform, too, it would be a seriously bad scenario (collision clipping). But think about it. Rotated images are only approximations on bitmaps. So what steps could I take to ensure that a rotated player coordinate doesn't approximate itself into the image?

Another thing I'm curious about is how costly doing collisions with rotated images is in non-HWA & in HWA, if theres any difference. Is the collision mask calculated each time it rotates and therefore no problem to use? Because that doesn't seem likely in HWA that uses render-level polygon manipulation.