Good question.

My thought process when I wrote that was that having the process already running wasn't an exceptional event. It might be noteworthy, but not worthy of a non-zero exit code. But that's entirely subjective to the use-case. I had recently been working on something where it was fairly normal behavior for an instance already to be running, and for that to be a perfectly acceptable situation.

So that's really it. A die would be fine too, but would (on a POSIX system) result in an 11 exit code, resource temporarily unavailable. By separating the message from the termination I went the other direction where the exit code is zero, while STDERR still gets a little quip.