I agree that other factors would prevent students having too much power, but "good" is a very subjective term. Even "effective" depends on how the students learn. Just because I think a teacher is good for challenging me, another student might hate them for being so tough.
Even if students are colluding to give poor ratings to teachers they dislike, it won't significantly alter the relative rankings of the teachers - the better the teacher, the fewer students that are willing to systematically exaggerate the teacher's weaknesses, and the best teachers (the ones who make learning fun) won't have their scores hurt at all. Students colluding will make it hard to set a threshold for "good enough" to not fire, but even if the students catch on, it should still be a good tool for identifying the best teachers.
I mean at the end of the day it's not really an opinion question; either the survey is empirically validated or its not. At least for now it appears to be, whether it will remain so when widely used is anyone's guess. (But not up for debate, since again it's something that can be tested empirically.)
