My Ph.D. thesis was based upon CFD. I made a point of validating my CFD solution against an exact closed form solution (almost no one does that). Good CFD can be extremely useful. Most CFD work is not good but almost always cheaper than experiment (Beagle 2 made extensive use of CFD results). Also even the crappiest CFD can usually produce pretty picture and colorful movies. In stark contrast, the best experiments typically produce only a few points on a 2-D plot with big error bars. No wonder the bean counters prefer CFD over wind tunnel data.

Gary