It would be handy to be able to add new keywords to existing language scanners, such as Java. This would allow users to leverage existing TokenMaker implementations with minimal fuss, assuming all they want is new keywords.
This would be accomplished by checking each normal identifier parsed against the "new keyword" list. If found, it's highlighted as a keyword.
One possible issue is performance, although there are TokenMakers in RSTA (WindowsBatchTokenMaker, UnixShellTokenMaker) that already work in the manner suggested above and don't seem to have performance issues.
Another thing to consider is making ALL keywords done in this manner (e.g. none hard-coded). This would allow users to replace a language scanner's keywords, not just append to it.
This same technique could also be applied to the "Function" token type as well as Keywords.