Element domain tokens help with standalone generalization

True. This is one of the original DITA 1.0 reasons for domain tokens.

But … generalization is rare, and already requires some knowledge of a document type. After all – how do you know what modules to generalize if you don't know what modules are present?

The token (topic pr-d) is meaningless if I don't already know pr-d is code for the programming domain.

If I know that, then I don't need the token to be present in @domains.
If I don't know that, then I have to start digging into my file. Finding the token (topic pr-d) doesn't help unless I know the @class attributes of the domain I wish to generalize. If I know the @class attributes, then they already provide everything I need to generalize.