I am dealing right now with a half marathon that had the opposite problem-- turned out to be about 0.35 mile too long. I've remeasured the course and am working on the map and paperwork. It turns out that the calibration course was about 43 feet longer than the measurer thought it was-- and the "overdistance" percentage in the cal course is a close match to the "overdistance" percentage in the course.
That's one good thing perhaps--it shows that the measurer was using decent technique, just based on the wrong numbers.
This is one I feel awful about because I signed off on the course and the cal course. Lessons I'm taking from this:
Measurer told me that her mile points were coming out consistently longer than the former points over the unchanged parts of the course. At that point I should have given her more options for double-checking! (try a different cal course or remeasure the one you're using, etc)
Always insist that measurers conduct some kind of double-check of their cal course distance-- the bike check is a good one. And as others have said, with experience we should know what sort of numbers to look for when calibrating, and anything outside of that should make us sit up and take notice.
I agree strongly with those who are reminding us of how crucial it is to use a number of redundant references (and a nail where possible) to identify calibration endpoints. More redundancy can help us avoid those mistakes.