While treating sunk costs as sunk is good, it's also important to avoid futile thrashing between alternatives, and to recognize that you have imperfect information at all times. There are times where it makes sense to continue to solve a problem the same way, vs. switching to a newer and better solution as soon as it is epsilon better. (generally, people have a lot bigger problem with treating sunk costs as sunk, but there are cases where they thrash back and forth and end up completing neither)