Right. I was skeptical in first few paragraphs, since students' teaching assessments do have their flaws. But if there was a statistically significant improvement in his students performance in later classes, then the man's doing his job.

Yes, there could still be confounding factors. He is seen as an enjoyable lecturer, and students who care about their classes enough to select the better lecturer might be better students. But still.

If you want to continue the analogy to software engineering, the modules he's producing are students, and they are meeting their specifications (later performance) quite well.