Do we assume only 1 ray passing every pixel? Can we also apply the supersampling idea here to improve the result?
Another question I have is that the ray passes the glass ball without changing the direction and later hits the triangle. Shouldn't there be refraction, or are there actually lights in all directions?
^You could process more rays per pixel, though it is probably not worth the cost. Ray tracing and rays are extremely expensive and doing more than is needed to get a decent image will just tax your GPU. Though there are probably some methods that try.
One question I had is how to deal with black body systems that absorb all forms of light, there are some cool distortion effects from space like simulators/movies/renders caused by blackholes so it would be interesting to see how to trace rays in situations like that that absorb all forms of light (especially the lens/distortion effect on the outside)
For your second question, I would say that if the ball depicted were glass, the direction would change, but I think we are assuming this is not the case right now. I don't think there are lights in all directions at the moment.