When converting PROHIBITSPELL on class level lines, the unparsed token gets output twice. See datachanges.txt extract below
Watching the processing, the token is parsed first via a WrappedClassToken instance (which calls ProhibitspellToken), but does not get unparsed by that instance. The processing then falls back to being processed directly by ProhibitspellToken but when it is unparsed the entries added by both parse instances are output.
Note: This issue would have affected any CLASS-only tokens appearing on a class level line for 1st level.