For example, one of the big projects in HIV research now is to get very good and very fine structure of viral proteins on the envelope of the virus; knowing which regions are readily accessible would make for potential good drug/cell/antibody targets.
Finally, the structure itself can give some insight as to function.
Most relate to the fact that protein function often depends on specific domains, and while a protein may have multiple functional domains it is important for all domains to be properly aligned and constructed in three-dimensional space.
Misfolded proteins often have negative phenotypes, so being able to visualize the improperly folded areas can be enlightening.
Interesting note: The Guardian (which is based in the UK where crystallography, protein or otherwise started out) has posted a short video outlining highlights of crystallography's contributions over the past 100 years.
People smarter than me can recognize those features and can learn quite a bit about a protein just from its shape.
Generally speaking, why do people spend all the money on synchrotrons, laboratories, robots and so on, for a crystal structure?
Protein structures, which can be obtained from protein crystals or from concentrated solutions of pure protein via NMR, are arguably the primary source of knowledge that we have about how genes perform their function on the molecular level.
Perhaps more commonly, these techniques can be used to validate an artificially produced protein before approving it for therapy.
Additionally, while we can know the sequence of amino acids in a protein very easily we don't necessarily know which ones are useful.
Having studied a little geology, I immediately thought of X-ray diffraction, but I could see how X-rays might be more problematic for proteins.