How do you troubleshoot a technical problem when your initial hypothesis is wrong at Los Alamos National Laboratory?