So it looks like the majority of our remaining failures are related to the language system. That's not surprising given that the locale system is not hooked into the new path handling at all. I suspect (hope?) that if we can get that wired up it will wipe out a whole bunch of failing tests. Getting that wired in, though, is non-trivial. I am unclear at the moment how much of it is core and how much of it nominally has to go through hooks. It's possible that we will implement a temporary listener that stubs out to a hook, just to make it work, but that will not be the final solution.