But this wasn't a different prompt that made it work better. Here they created a database of prior human created comments with dev ratings of how good the devs found them.

And then they fed that database to the model through prompts. That's prompt engineering.

Is there a name for the activity of trying out different strategies to improve the model? Before one prompts it in the wild.

I'm not sure, I'm not really an expert in the field, though I do this professionally (but it's just a tiny part of what I do). I just call it A/B testing...