This is a tricky area perhaps best left to the lawyers. But even by the free software foundation (FSF) strict interpretation of the GPL, this sort of thing is allowed. S is a programming language, and R is an interpreter of that language. The FSF foundation then argues that the program is just data with regards to the interpreter, and thus constitutes a separate work (from a copyright perspective). Therefore, it is not bound by the GPL. see: http://www.gnu.org/licenses/gpl-faq.html#IfInterpreterIsGPL
The R core team concurs with this, and at one point released a statement on R-dev that alternative licenses for R packages are allowed. In fact on CRAN you will see a wide variety of licensing terms, some of which are not 'free' (i.e. optmatch).
It is another matter if Revolution's R packages depend on other (non-base) packages released under GPL (I don't know if they do). The FSF would argue that the packages would be bound by the GPL, but other legal scholars would disagree. I asked Stallman at the UseR! conference whether he thought it was likely that this would be tested in a court of law anytime soon, which is the only way to settle this sort of thing, and he said no.