Whenever I have run into this kind of problem in the past it usually gets back to a difference in the actual chemistry or treatment of the steel. You might submit a sample to an independent lab for analysis. If you have any of the of older more reliable screws around, you might submit one of those at the same time for comparison. You could also try samples from a different supplier. Can you set up a test sequence to simulate the conditions to try to duplicate the failure internally?
I have some very painful memories of watching specially built samples of trailer jacks fail miserably one after the other at a potential customer's site after all our internal trials had been very successful. Turned out the customer test units were made from a different heat of the material at the steel mill.