I scraped google API exactly like this in grad school, but sometimes when you're building, you take a calculated risk to build something better.
For my thesis, I was trying to load two Street View photos side by side in a browser to compare people's perceptions side-by-side. Google maps at the time required you to load a javascript viewer for each image you requested. Think Hot-or-Not for cities.
The experience needed to instantly load a new image after the user voted because I knew they were only going to be on the site for maybe 30 seconds before they got bored and went back to reddit. I needed to collect as many votes as possible within that time period.
So I knowingly broke Google's ToS and prefetched the images on my server so I could provide the user experience I wanted.
"I'm a small operation. Surely they won't know." I wrote a server side screen scraper to load the Street View images and exposed the scraped images with an API.
Now my site was faaaast. I could load Street View images instantly and in the end got over a million data points doing this.
But then one day soon after it stopped working. Then I got a cease and desist email from someone at google legal. They didn't respond to any requests for turning it back on or even to discuss. Radio silence. That was terrifying.
Since this was my thesis, I needed help getting my keys turned back on. Google in the end was very accommodating, but only after I used my nuclear option: asking lab director Joi Ito to bug Megan Smith while she was still at Google to help.
I was connected to some engineer and told them what I was doing and why. They said stop. But then a week or so later, they sent me a beta invite to their new Street View images api, where you can feed in a lat, lng, header to a query string and they'll just serve the image now. Pretty cool.
Offending site for the interested: http://pulse.media.mit.edu/