From what I understand (given this previous question and the Google Maps API docs), it seems that an attacker could run up another site's quota.
For example, let's say I run a web app, "Find Taxi", and a competitor runs "Find Ride". In my web app's JavaScript code, I include my Google Maps v3 API key. The developer for "Find Ride" runs a PhantomJS process that spoofs the referrer header (as http://findtaxi.com/) and loads 25,000 maps per day on his server (running out my quota limit). Note that even given the client IP rate limit, 1 request/second/user, the quota can be reached from a single client within one day (86,400 seconds in a day).
I created a simple PhantomJS script to simulate this (and an API key restricting the referrer to example.com/*), so this attack appears to be feasible. Is Google relying on non-technical reasons to prevent this? For example, I can see legal issues (the threat to the business if this attack was leaked) and economic issues (the cost of running the attack server).
I guess I'm asking this question in case I missed a technical reason why this attack isn't possible.