The Eclipse Mylyn project has been promoted to a top level Eclipse project under the banner of Application Lifecycle Management tools (though the Mylyn name is being kept as a short name). There is a project charter which explains its purpose in the ecosystem.
What this means for the average Eclipse Mylyn user is (directly) very little difference. Mylyn will still continue to be developed by the same set of committers as before, and provide the same kinds of developer-focussed experience that is in use today. Being an Eclipse top-level project is really a symptom of Eclipse's anachronistic project structure at the moment, which has a two-level enforced structure for organisation of projects, like the Eclipse project's JDT and PDE, and the Tools project's CDT and PDT
Mylyn has taken some time to find its home, having come through an incubation phase, through the Technology and then Tools projects, until it's final resting place at the top-level. (The majority of projects are either in the Technology or Tools projects, and the choice of which is both arbitrary and in most cases completely invisible to the users and committers of the project.) Primarily, the split is based on the CVS or SVN root, which will become moot as projects migrate to Git and throw away the shackles of the enforced repository structure.
As part of the reorganisation, Mylyn is creating some second-level projects of its own. This is really just a repartitioning of already existing Mylyn modules, with the EGit and CVS connectors being part of the Mylyn/SCM project, Bugzilla and Trac connectors as part of the Mylyn/Tasks project and Mylyn/Context, which forms the basis of the active UI filtering that brought Mylyn its initial popularity.
However, Mylyn hasn't remained static over the years and is diversifying. The Mylyn WikiText component, originally for writing bug reports in a friendly fashion, has been spun off to a separate Mylyn/Docs project, where it joins forces with a RichText based editing cousin. Similarly, Mylyn is extending its reach to integration with Hudson, which includes the ability to read console output and hook into failed tests on the server as easily as a local test failures.
Finally, Mylyn is setting itself a goal of providing an interface to review-based systems, with initial plans to provide a simple Mylyn Task-based review system, as well as hooking into existing review systems.
For more information, check out the interview with Mik Kersten and the project charter.