Thanks for your suggestion, but...

The example currently in the docs is exactly how we do deferred translation in the project I am currently working on.  Your example is much more complex, and I don't see the benefit of it.

Specifically, using the example in the docs you can easily mix deferred translations with non-deferred translations, which what you generally end up wanting to do in real code.  (You use deferred translation with strings that are defined at import time, as in the example, and non-deferred with strings that are computed at runtime.)

Perhaps there is a need for additional text about how this works in practice, including an example of switching the language dynamically during runtime.