tl;dr - By doing a detailed analysis of his app, he was able to discover that a lot of what he was using App Engine for wasn't being used or could be made more efficient with relatively little effort.
Having worked on App Engine, I'm sure almost all (if not all) of the apps I've built ended up in this situation. Because it was "free" there wasn't a lot of reason to focus on being efficient (especially in terms of space). Getting an app to work and serve its purpose while maintaining a level of perceived performance was the top priority. Any work after that was usually not worth it because the apps tended to be small in scope and non-critical.
That said, it feels like if I were to build on App Engine again, it would require more development time to keep it free. With a couple of scripts to setup a free micro EC2 instance, that might make it easier to build those small apps there instead. Not sure if that's good or bad for App Engine, but I suspect bad.
Could someone please explain to me why this is different from hosting your own environment? In your own environment, $x / month gives you an environment that can handle a certain amount of load. You can either spend time (= money) on optimizizing your application so you get more bang for your buck from your hosting environment, or you can spend that money on your monthly hosting bill. And it's your job figuring out an optimal balance between those two things.
The advantages of appengine compared to an EC2 micro instance are that there is every little that can break over time. Develop once, deploy, forget about. Also: It's only free the first year, you still have to pay for bandwidth, it will cost more effort to scale and you can just have 1 free micro instance, but 10 free apps.
Having worked on App Engine, I'm sure almost all (if not all) of the apps I've built ended up in this situation. Because it was "free" there wasn't a lot of reason to focus on being efficient (especially in terms of space). Getting an app to work and serve its purpose while maintaining a level of perceived performance was the top priority. Any work after that was usually not worth it because the apps tended to be small in scope and non-critical.
That said, it feels like if I were to build on App Engine again, it would require more development time to keep it free. With a couple of scripts to setup a free micro EC2 instance, that might make it easier to build those small apps there instead. Not sure if that's good or bad for App Engine, but I suspect bad.