Here is a suggestion for the Aurora team based on my experience. Having been both a rocket scientist and a bio-med engineer I believe that the industry methodologies that will take you where you want to be are in the bio-med engineering area not NASA or the space program. I worked in the diagnostic industry for J&J and one of my projects was to get the Troponin assay to a false positive rate of less than 1 in a billion. The space program and rocket technology has a very small data set and a relatively high failure rate. Commercial aviation is a better model industry to model but their system complexity is lower than bio-med.
I was able to change the failure rate from 1 in a 100 to over 1 in a billion in three years. The key was to both improve the processes and the failure detection (after learning the failure modes). I read your blog and many of the things you are doing are what I did with my team but I don’t see the concentration on failure detection that may be necessary. I have a number of published patents that demonstrate these methods http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch-bool.html&r=0&f=S&l=50&TERM1=Merrit&FIELD1=INNM&co1=AND&TERM2=Jacobs&FIELD2=INNM&d=PTXT