Actually I'm kinda hoping its a little of all of the above, Except there should be consequences for some choices, Going one way Precludes another way, and you just have to make it work. To me The Ideal research schema would go some thing like this Traditional research - Bigger, Better, Faster. Improvements to your Rockets and various Components, New types of scientific equipment, etc. Solely Money based (Actually a side Idea to this is similar to what was said before, except it uses the "Corporations" that manufacture the parts, the more parts you buy from them the more "Stuff" they can develop to sell to you, Perhaps Making them an Official supplier reduces the costs of their parts but increases the cost from others.) Special Projects - As you explore the system, and encounter new environments and materials you Unlock projects where you can get a bonus to all craft or with mining make your own parts, or even open up new "traditional" Research Subjects, in areas like Alternative power generation, and Engine Fuels. (These would require actually acquiring said materials to work). Scientific Projects - (Maybe random when they appear based on condition on Kerbal) Example a Scientist has convinced the Scientific Community that there is life on Duna, If you send a Mission to Duna to prove either way, you get some bonus (Perhaps in the above example you get a bonus based on the result of the test) I'm thinking these should have a time limit on them, as "Scientific Fads" only hold the public's attention for so long.
