The underlying problem they’re trying to detect is acquiring a resource in a thread before forking that is then held forever after forking (because the acquiring thread is gone, and so cannot release the resource), and then attempting to acquire the same resource in the forked process.