When i was an automotive powertrains man we used to use an exponentially decaying curve to estimate the expected number of faults per test hour. It is suprisingly true in other areas of engineering. The problems never go away, but as more "unknowns" reveal themselves the development team can identify and solve the root cause.
Given the statistical nature of this development process, it always amazes me that it is generally the money men who struggle to get it...
Mart