Best practices when using image map files?

[Horbgorbler]

Question:
What are some best practices for using image maps in MB3D?  I have found that I can store .jpeg or .bmp or .png files in the maps folder and that the pixel dimensions of these files can vary widely.  For example the image can be 256 x 256 or 1600 x 1200 pixels or even larger.  But I am not sure what the tradeoffs and consequences really are.  How important is it to prework the images to tile seamlessly? I see that the identical image files can be used for both light maps and height maps.  Does MB3D convert color images to B&W gradients when they are used as a height map?  Is there any advantage to having the initial file B&W as opposed to color?
I know that all of this depends greatly on what kind of surface modification effect the artist may ultimately be after.  But I have no feeling right now about the relationship of image dimensions or pixel sizes to rendering time.  Is a smaller image dimension or file size generally better?  Are square images generally better than rectangular or perhaps masked circular ones?  I see a lot of discussion about bit sizes and excitement over 48 bit images but am too ignorant to grasp why and when this matters. 
I have just begun exploring the use of height maps in the IFS formulas provided in MB3D and am amazed by the effects that can occur. But I know that I am missing a lot.
I really like the concept of using fractal images for the height and color maps so that the 3D geometry is enhanced even further via fractal patterns of surface texture and light.  That approach takes the fractality to an even more pervasive level.   If someone who has thought through these issues could comment, I would greatly appreciate it.  In the meantime, of course, I will continue to happily experiment and play.  Thanks in advance for any feedback.