I have one computer that's apparently poorly suited for CP and one that's well suited for CP, yet I want to support CP, Einstein, and LHC equally.

But if one computer runs all three and the other runs LHC and Einstein, my global resource share would not only be a function of which computers are attached to which project, but how fast each computer is (which is impossible to calculate with complete accuracy).

Wouldn't it be much easier to control your resource shares globally, so when BOINC decides what to download and/or crunch next, it looks just at the total credits for each project and picks the one with the largest deficit with respect to the global resource share?