I had actually thought this might be kinda obvious. But I got a heartfelt request, so here goes. I wrote:
It is a sign of maturity to let go of people when it is best for them. It is a sign of maturity to know when it is best for them.
And someone said: but how do I know, really know? Of course, letting them go is not taking up scissors and cutting a line. Its getting a postdoc to apply for jobs. Its helping them sort through the offers they will (of course) get. Its directing them to Doc Becca's aggregator page.
First off, you are not immature if you Don't Know when to let go. Its knowing and not doing that is a problem. How to know?
Signs a postdoc is ready to go:
1. They have the ovaries to question you, but aren't doing it reflexively. Or, they never questioned in the beginning and now they do. Or, they always questioned you and confronted you, and now they have stopped.
2. Their ideas are ones you wished you had. They understand what is going on in the lab and are thinking of new directions. This is a moment of danger, because the temptation to keep them and their ideas is large.
3. They take the lead in journal club, when they didn't before. OR... they let others talk, when, in the past, they were always the first to talk.
4. They start catching little mistakes you (of course you do) make. And, they are gracious about it. This is also a danger point - their value as a colleague has just doubled.
5. You don't have to oversee their mentoring of younger trainees. You realize that the way they are working with younger trainees is something you admire.
These are diagnostic signs. They are not Truth and Beauty. They do not all apply to any or everyone. They are not absolutely necessary, and may even not be sufficient. But they can be guides.
The hard part of knowing when to let go of a trainee who does not have a ticking clock (the way a UG does) is the realization that they are not you. They will not follow the path you did, even if that is possible. A good mentor recognizes this. We often make type 1 errors (letting someone go when they're not ready), and that can be bad, but is seldom fatal for either you or the trainee. But type 2 errors, poignant as ever (keeping them beyond the time to let them go), can be devastating.