There’s a long history of tech companies developing there own applications because it’s cheaper long-term than licensing, especially for core applications like: email, calendaring, text processing.
I’m confident Apple employee use Mail.app, iCal, and iWork in-house and those apps are cheap or free for the rest of use. Same for Sun and StarOffice/NeoOffice. Same for Google and Google Docs. One guess on who isn’t getting ongoing licensing fees for those apps? 
This is not only why ‘HuddleChat’ was the first AppEngine app but also why it was pulled. HuddleChat made it obvious.
I can easily imagine this conversation at Google:
“Campfire is a great tool, we should pay for it.”
“That sounds like a lot of money for something not built here. It’ll be cheaper long term if we build a clone in-house.”
Take a look at the number of applications in Google’s Lab page. Many of them need; some form of authentication, the general look/feel of Google, integration into Googles infrastructure, to be built at Google, etc.
What a perfect candidate for an abstracted framework like Google AppEngine.
Elsewhere:
Confirming this theory and that AppEngine is all about future acquisitions (i.e. ‘Want to increase the chances of being acquired by Google – build on AppEngine’).
Interesting take on the GAE. I like the “scratch our own itch” theory you’ve come up with.
How do you know that Google did GAE for themselves? It’s a company that is organized much differently than the other companies you mention-so I wouldn’t assume that’s what this is.
As far as the quotes in your post, I guess time will tell if GAE has staying power.
Jen – this declaration is based on my readings of the struggles and challenges acquired projects have in getting integrated into Google’s infrastructure
http://www.techcrunchit.com/2008/07/16/google-where-companies-go-to-die/