Problem is that sequence homology over a small tract of a big macromolecule does not guarantee structural homology, as I'm sure you also know.
The COG CDD projects at NCBI (among others) are trying to address and catalogue various shared structures and domains in proteins.
Domains and motifs can be mixed and matched in different proteins. Recombination events can lead to oncogenic states when domains are mixed. An example is the Hoxd11 were were discussing earlier. The idea that that gene is somehow specific to or in itself directs limb morphogenesis is not really right and I'd hoped to bring that up at some time and this works out well. Hoxd11 as are so many Hox genes (as well as hedgehogs, also mentioned in the SciAm article of another thread) are involved in development seemingly rather generically.
A fusion protein fromed by recombination of a gene called NUP98 and Hoxd11 is, for example, a cause of leukemia. Hoxd11 is found in hematopoietic stem cells and when it becomes dysregulated by the fusion even, the stem cells do not develop (or differentiate) correctly, leading to a proliferation of leukemic cells.