When research goes off the rails: why it happens and what you can do about it