You run split experiments over subsections of the line to find the source. It's essentially like binary search. You can do this pretty quickly with a skilled team.