At one stage we increased the time between failures of F27 gyros by simply not changing them as often - they were extremely sensitive to shock and we were able to prove that they were often damaged (brinelling the bearings) in the process of fitting them, a rather extreme and expensive case of infant mortality.
System reliability tends to show a bathtub curve. Most errors in assembly or grossly dodgy components tend to show up quickly. Once they are repaired/replaced the system settles down to it's expected level of reliability until it starts to wear out and defects appear at the end of its life. But of course we are talking large numbers of the same system here.
I keep asking engineers to go with me on first flight after maintenance, for some reason, they always decline.
