Not really workable in practice. Are you going to lie in the project schedule section of the application? Nowadays the grant agencies expect to have Gantt charts and other crap. It also would require delaying publishing a work when it's finished, which then raises the chances of perishing...

What people however do is to smuggle more risky ideas into proposals for less risky ones. But that in the end means they'll be spending time also on low-risk things.

Probably works in mathematics and other theoretical sciences. If the way you use your funding depends on the particulars of what you're doing, then you're probably screwed.