I don't think the decision had anything to do with money. It's part of the whole "more wood behind fewer arrows" initiative. Essentially, a defrag. 20% time was seen as having created too much fragmentation with little benefit. Gmail was the only major success to come from the program. Also, Google has a separate team for working on skunk works style projects.
Metaphor is not reality, but that metaphor is terrible. I'm an archer: you want as many arrows as fit in the quiver, and as far as more wood, well, we've switched to carbon fiber shafts for a reason. When you shoot an arrow, velocity gets you to your target, and velocity penetrates it too. Extra weight is a literal drag.
While Go doesn't directly generate any revenue, it's probably not without reason that Google is rewriting some important pieces (dl.google.com for example) in Go. I can imagine their engineers being much more expressive in Go than in C++, causing them to spend less time on writing the code as well as resulting in simpler code (easier to comprehend and maintain, so less time/money spent on maintenance). This doesn't generate revenue but it does optimize development processes, which could indirectly save them some money.