Thanks Wolfgang!
It's solved, your input put me on the right track. The user who designed and started this job had started the job manually. The job is actually a "child" itself, in normal operation it has a parent. The parent activates a "job group" that was bound to the child job.
When he starts the child on it's own, he needed to start the job group by hand, too. This is what he initially neglected. Only then does the job start to run.
I guess a proper message like "waiting for job group blahfasel to become active" would be an "Ideation" item ...
Best,
Carsten