It's like chicken-egg situation. Start using go for non trivial parts. But as you move ahead, you will find that some things are close to impossible You drop to C, and end up using C++.
hm... honestly I can't see what's missing in Go to force you to drop to C. .. it's absolutely usable for a simple game without expecting bad surprises... most game oriented bindings are already done.. I seriously can't think about a single thing un-doable with Go.
Using it for a big game tho, that is a jump into the unknown at the moment, we need to get there step by step.
It's also important Google keeps this project going. Go 1.1 is expected to hit the internet next year, afaik it'll be mostly about runtime performances.. if the Go guys can deliver this, I think things might get really interesting.