I just found about the extension today from Dave via IRC conversation actually ... your suggestions sound great!
So if we are thinking about FEATURE testing for example, which for the most part is little more than user interface testing of a particular package, I would imagine that what we would then end up with is the following workflow:
1. Adopt a package or group of packages for testing.
2. Develop selenium test scripts.
3. Post test scripts links on the package pages (I believe the wiki has the ability to link to uploaded files, I will talk to Dave about this).
4. Use scripts as repeatable test cases on each bug fix iteration.
Would that be how you see this working?