Following up to my last post, it turns out that CSALTER requires you to have clean data and produce a clean CSSCAN within 7 days prior to running the character set migration script.

If the CSSCAN isn't clean, you'll get this:
Checking data validility...
Exceptional data found in scanner result

PL/SQL procedure successfully completed.

Checking or Converting phrase did not finish successfully
No database (national) character set will be altered
CSALTER finished unsuccessfully.
So that question has been answered. Now to scrub the application data, as well as the histograms for that data that are in the data dictionary.