Let's back up a minute because there are some inaccuracies in this thread.
ai2-test.appinventor.mit.edu is the test server with advanced access to new features and is used for testing functionality and bug fixes before they are released to the main servers. It runs on the same infrastructure as ai2.appinventor.mit.edu (Google App Engine). When we prepare a release, we first update ai2-test and then after we do internal testing and a period of time with no bug reports from other users, we push it to ai2. Functionally, the main difference between the two is that ai2 has the gallery functionality enabled and ai2-test does not.
code.appinventor.mit.edu runs on an entirely separate infrastructure (Amazon Web Services). We consider it a production service at the same level as ai2. It mainly differs by 1. offering an anonymous usage mode for Hour of Code (from which it derives its name); 2. no gallery; 3. larger build sizes; 4. accessible in China (due to not being in the Google cloud). We update code usually within a week of a release to ai2.
Each service is its own system with its own backend. Projects are not shared between them. You have to manually migrate projects from one to another if, e.g., you plan to test a feature on ai2-test.