ADE Assembler: Global Variable Sweeps vs. Corner Sweeps, Practicality and Philosophy?

Hi,

I admit here that my subject breaks one of the forum guidelines, that questions be as specific as possible. I apologize in advance.

Are there any definitive reasons to perform ADE Assembler sweeps in the Global Variables (or Design Variables) vs, in Corners vs. partitioning them across both?  

1. Practical? 
Certain capabilities seem to work better (or work only) in one or the other sweep method. It would be helpful to have a table that compared and contrast those capabilities.
Also, it may depend on what variations I particularly want to group, but I am looking for the greatest general flexibility with the least rewriting of tests or duplications.

2. Philosophical?
To Cadence employees: Is there any philosophy of usage distinction that guides development and keeps the sweep capabilities a bit different?  Or is it just that both grew into similar functions, but started from different code? Or something else?
To super-users: Do you have a clear philosophy of how you partition between Global Variable Sweeps and Corner Sweeps? Are there practical reasons you do so, e.g., it makes your plots easier to obtain or makes your measurement easier to express.

I could go to Cadence Support and try to develop a specific solution to a specific design, but I am instead seeking a discussion of useful ways to organize this across many designs.

Thanks.