I follow this procedure before using any new software that may mess with my tags.
Make a copy of a few representative files, those that are pretty thoroughly tagged to my personal standards. Use these to test the software/script whatever, then compare the before and after versions to see what's changed.
In extreme cases I'll use the "Complete Tag Export" add-on to CSV files and tools like CSVfix that let me "diff" the data changes, but usually I can just eyeball the changes in MP3Tag itself.
For affected fields I want to protect my own data, I create custom tags for backup purposes, and use actions to copy the main tag values to these backup fields before running the update tool, and then another action set to copy thosed backed up values back to the main fields afterwards.
Of course I'm using the same tool to make complete backups of my meta data to text files anyway, and I happen to import/export/store this data in a separate database, so you can seem I'm a bit of a "belt and suspenders" kind of guy.
This lets me fully test and use whatever wild-ass data collection tools are out there, let them write to whatever they want, while MP3Tag lets me put everything back to my personal standards when I'm done with them.
For example, I've been expanding my canonical list of Genres to use, at the Artist, Album and Track levels, via LastFM Tag Extractor, which collects the cloud-source data according to very flexible rules into the Comment and ContentGroup tags as well as Genre, and the above workflow lets me set those rules wide open, writing different patterns to all three fields to collect the data I want, and then put everything back afterwards and tag my music manually once I've standardized on the genre strings I want to use.
It's actually very entertaining seeing what kind of weird folksonomy tags people will use for tagging music genres out there, and the process has given me great ideas for categorizing my music in dimensions other than the standard ways.