I guess I would ask what exactly these bonuses are targeted toward: teachers that students like, or teachers whose students have good results -- with the caveats about student performance as noted in the first comment, or teachers whose students show the greatest *increase* in performance under the teacher, or something else?

I would posit that the first is what is rewarded, at least at higher levels of education where a university department might solicit student opinion on professors: the appearance of success *is* success. Of course this can backfire when a teacher is well-liked because they are easy and not because they are effective and interesting.

As noted in Jeff's comments, the second may be beyond the teacher's control in many instances: where students come from fragmented backgrounds, etc.

I think the third is probably the most fair: if you normalize against performance at the beginning of, e.g., a school year, then you should be able to compare performance increases in many areas.

I don't have any real thoughts about measuring instructional difficulty in content areas.

Of course, I am a student and not a teacher, so take it all with a grain of salt.