Since tracking markers isn't really a supported feature, I don't have that information readily available. You should probably experiment to find the best results. It is more a matter of brightness then size. A bright, small marker could contribute the same amount of photons to the pixel in an image as a larger, dimmer marker.
Size is important to resolving two different markers, though. The angle of view of each camera is about 140x130 degrees (you can get the exact angles from the Device class) and the image size is 640x240. Thus, on average, each pixel covers .4 degrees (simplifying a bit). At 500mm, the area covered by one pixel is about 7mm (2 * 500mm * tan(.4)). You wouldn't be able to resolve different markers closer than this at 500mm.
The lenses exhibit complex distortion, but again, this matters more to resolving power and the accuracy of triangulation then on how large a marker needs to be to be tracked. For accurate tracking, you must remove distortion using the calibration map (or warp() function) anyway.