Wind Tunnels Are Not Perfect
Much has been made of the CFD-only strategy recently abandoned by the Virgin Racing Formula 1 team as an admission that CFD is responsible for the team's poor performance. Whether this is in fact the case is open for debate. I believe VR's poor performance is due to a general lack of funding across many disciplines not just aerodynamics, but that's another story. Anyway, as a counter to those who think wind tunnel testing is essentially flawless, consider the recent problems Ferrari have experienced with their wind tunnel.
In the early 2011 F1 races, Ferrari noticed that the aerodynamic performance (especially downforce) of their racecars was not reflective of the performance predicted in its wind tunnel. In search of a remedy Ferrari benchmarked its wind tunnel against Toyota's ex-F1 wind tunnel. It turned out that there was a problem with the Ferrari wind tunnel, which they claim to have now fixed. However, the full 'fix' won't be complete until October 2011, so I presume that the remedy was not trivial.
It just goes to show that wind tunnels are not always the custodians of aerodynamic perfection that some believe. Funds permitting, it's always a good idea to have a reality check to give confidence that the results being produced are valid. In the case of a wind tunnel, CFD is a great compliment to provide extra confidence in wind tunnel results. Of course it goes the other way too, i.e., a wind tunnel can be used to gain confidence in CFD results.