If the tolerances are right there shouldn't be any harmonics that would cause failure although I would put this as my second most likely guess.
I have been wondering if it's possible that cavitation could be the cause and if so then how much could be done to mitigate the problem. I know high RPM motor builders already port the pump to minimize this but could any more be done and if so would it help?
Most people don't think of cavitation as a destructive force but if it's extreme enough it definitely can be.
http://www.engineeringtoolbox.com/cavitation-d_407.html
I have been thinking about trying one of my dry film ceramic coatings on pump gears but I would need a controlled environment to test it in. Even then it would take a number of comparative tests to get any useful data. It would tighten tolerances, lower friction and give it better protection.