While I can't speak to the issue of whether computer testing should replace physical testing, what about the issue of design/test specifications? For example, wings are required to handle loads that are 1.5 times their typical expected load. Who chose the 1.5 figure? The fact that it is such a round number makes it suspect by itself. Indeed, over the past 60 years, how many planes have had their wings snap off in situations and attitudes that were otherwise recoverable? I think the answer is few to none. Thus, my hypothetical question: why should the number be 1.5? Maybe 1.4361 is more appropriate and economical? And while computer modeling may be less exact than physical testing, perhaps its margin of error is less than the difference between 1.5 and 1.4361 -- so using computer modeling and keeping 1.5 still results in appropriate safety.
Just my two cents.