I'd guess there is something like 50% cutoff point.

Less than 50% of hires stay: "oh geez, more meat to the grinder."

More than 50% of hires stay: "A new guy came, do you play ping pong?"

But my cut off point might be way off. And there might be some elasticity between those where people are more iffy about the whole deal.

I'd suggest testing this. Stack slightly not so promising new hires to one team and more promising in other. Let's hope there is different retain rate. Now you can measure the performance changes of both teams and see the net effect.