I don't think this is the best approach to take. The problem with this approach is that you're tightly coupling you application with the infrastructure in a way. Domain and subdomain are just an easier way to access a machine bound to a specific ip address. I would classify domain names to be part of the infrastructure, not really part of the application. If you go with the above mentioned approach you're introducing some knowledge about the infrastructure into the application and thus making your application less flexible. What happens if you, for some reason, sometime in the future decide that your client A should use clientA.mydomain.com? or how about keyClientA.myotherdomain.com? Or how about you want to allow your client A to use their domain name, i.e. support.clientA.com? If your application does not know anything about domains and infrastructure setup then it's a lot easier to just reconfigure DNS server and get the portability.
If I had this scenario, I would have some kind of mapping of URLs to a tenant id, and then use that tenant id as a namespace name. This way you can easily map different URLs to a tenant id. You can even map multiple URLs to the same tenant id and expose the same application on multiple URLs. Your mapping can be stored in a simple config file or even in the AppEngine datastore itself within global namespace. If the config is stored in the AppEngine datastore, you can have your admin section of the application (or even another AppEngine module) which you can use to update config in real time.