This is very similar to what we do in Logger, though we do it more explicitly.
Logger package instances and logger projects are independent, and are linked with a mapping table.
The way we use it internally is this:
Mounted at our intranet, we have a logger instance, which is mapped to all projects in the system.
Mounted at each of our client extranets is a logger instance, which is mapped to the projects involving that particular client.
I could see the use of a similar model for bug-tracker.
/Lars
