paull;
If this is being done properly then for any fault (in the larger "man-machine sense") Airbus or Boeing would be able to immediately quantify the impact and potential frequency and use this as a process for driving continous improvement. Please tell me it is already in place.
In a word, yes.
As you have seen by the responses already posted, there are organized "audit" programs in place called FOQA (Flight Operations Quality Assurance), or FDA (Flight Data Analysis).
With some programs, the de-identified data (no date, no flight #) is kept instead of destroyed, so that questions can be asked of the massive (in the terabyte range) database. This has been mentioned in the thread already. The technique is to construct a question (using PERL or other query language) to "trawl" (good word) through the data. Ordinarily it would take weeks with such a huge database but there are ways of doing this quickly.
in the recent Airbus overrun there is much discussion about just how unusual it is to forget to pull the throttle back on landing. Well how often does it happen? It seems to me if we want to systematically drive improvement then you need to know the answers to these questions without waiting for people to die (a few times).
We are already now asking this question of our stored data.