NASA, for example, does not allow grant funding to be used to write grants. So, this preliminary data thing sounds like a different model. Where did the money come from to obtain the preliminary data? With regard to NASA, grants can cover administrative overhead. And, most institutions have support for new grant writing efforts. Doubtless, some NASA grant money that goes to overhead ends up providing support for that kind of effort so new grants do get written. It is just murky.
In any case, all that work to find out if an idea is technically feasible enough to make a good grant proposal gets paid for somehow to persuade peers that a proposal is viable. So, really, the originality of new grant proposals has something to do with how well faculty are supported in exploring new ideas. That would seem to be the place to ensure that peer reviews get to see exciting and not just competent proposals. Are the institutions hiring the most creative postdocs, for example? Are junior faculty getting good seed money? Is there time set aside for use of laboratories for pursuit of hunches? So, if granting institutions want to see more creative proposals, they'll have to look at the institutional culture grant overhead supports.