Google App Engine's pricing model costs me way more than the small amount of money I'm charged every week. It costs hours of time.
Now what do I mean by that. I love App Engine and use it every day, there is currently no faster way to get an experiment up and running. The pricing model sounds great: Pay only for what you use. But without realizing, that this model can take giant tolls on development time, it will. At least if you have to care about money.
The problem develops as follows:
1. You have a great new feature to implement.
2. You design a rough implementation, that will get it working asap.
3. You notice a few unnecessary datastore/bandwith/cpu calls/bytes/hours (which AppEngine charges for), that might cost you thousands, if the first few million users hit your site on launch day...
4. But no worries, this can be solved with a slightly redesigned data-model. Go back to 2.
All the momentum you can build up with GAE (setting up a new project in about 5 minutes) can be destroyed in this vicious circle.
Maybe this is common knowledge, but if you are like me, try this:
Ignore everything you know about GAE optimizations and get your feature implemented.
If, one day, millions of people should hit on your site, be happy, pay a higher bill that month and start optimizing.
This should apply to a lot of "Pay as you go" - models out there.
Again, maybe this is just me learning obvious stuff, but this mindset saved me hours, and saving time is great.