As I understand it, its main feature is the ability to test applications which use double-byte character sets such as Kanjii, Mandarin, etc.
I've seen it. Interesting to see Kanjii characters as tags and being able to use double-byte characters in verification. Very helpful if testing any such site as prior to this you had to use Index tags for everything and just take it on faith that it had the right text.
Of course, since I can't read Kanjii, Mandarin, et.al. I guess I'll still be taking it on faith that the characters are correct.
In Winnt/2k, the OS kernel is UNICODE, and the OS will translate most of the WM_xxx from DBCS to UNICODE. So if your application
1. doesn't require ExtensionKit,
2. or your application uses the common control provided by Microsoft,
3. or your controls can handle UNICODE version messages,
you can use STi free of headache.