Given my background in stats, I admit to a tendency to classify variables.
So… first you have to decide which, for you, are deal breaking variables. I’d include trainees (plural) that talk trash about the mentor, vague hand-waving when talking about projects, people who don’t return your call/email (ever).
Then there are variables that get thrown into the hopper and balanced out: what you think of the projects, how many people in the lab, are projects shared, what are former trainees doing, age of mentor, etc.
When I decided to go to grad school, I narrowed it down to two rather different situations (although the PI’s were long time collaborators). I made a list of what was important and gave each factor a weight between 0 & 1. I think I had a list of about 30. Then separately (on other days), I sat down and ranked both places on each variable (without looking at values of variables). I did the calculation. They came out too close to be statistically meaningful. So I flipped a coin. I swear I decided where to go to grad school on a coin flip. No regrets about what transpired.