Because it might be the best option for a next few years? I'm not saying it is, but some people did a right choice "locking" themselves to Adobe-only infrastructure (Flash), Apple-only infrastructure (iOS/Objective-C), Microsoft-only infrastructure (Windows/Office/.NET), ...

There's no real lock-in as long as you cushion your application code from their APIs and infrastructure reasonably well. I was worried about lock-in myself with a side-project I took on a year ago. I decided to go with with GAE despite experience with raw AWS, just because I was a one-man-team about an hour a day, and didn't have time for all the technical overhead. But I made sure that I could switch to more flexible hosting without much difficulty. Their Datastore might be the hardest to abstract away if you rely on it too much, so I wrapped it inside a non-queryable blob API that could really have any kind of backend to it.

Discontinuing Reader was one thing, but there are certain services that Google couldn't discontinue without it irreparably damaging the company among either consumers or business. Reader was close to being that, but at the end of the day no ones business depended on Google Reader, and mainstream consumers barely read Google News. Reader wasn't mainstream.

App Engine? If Google ever wanted to discontinue it, they would likely spin it off in some way. A company like Rackspace would surely be glad to take it off their hands.

Because you can host your App Engine applications on other platforms yourself?

http://www.jboss.org/capedwarf

http://www.appscale.com/

Also, I recommend developing for App Engine as an educational experience. Building a good App Engine application may seem a little bit constrained, but it teaches you a lot about how to make something fault-tolerant and able to scale horizontally.