On theoretical and practical considerations of path selection for delay fault testing