Do you think the grant could go directly to fund research, with the university then funding a grad student with the grant money as a condition of how the grant is used? Or does it have to go through the grad student as a reportable tuition waiver?
I certainly do agree this sounds like a much more complicated problem than a line item that can be hidden when things are internal. Just wondering what loopholes might work.
My impression is that the humanities are a bit different: grad students are funded by the university, but also expected to do more TAing (etc).