Here is a way to perform a QA for in-project target term consistency, which is based on the assumption that segments that only contain one word (either at the source or at the target, or both) likely contain your preferred target terms.
Of course, this assumption doesn't have to apply to your projects ...
Here is a movie: https://youtu.be/u1PA2ps0j48
Most likely, you'll have to fiddle around with the regular expressions for your languages. Note that \A and \z indicate segment start and end, respectively.
You can try to further optimise these regular expressions, e.g. for German you could say that you only want to check for nouns and add a criterium to the regular expression that the segment must start with an uppercase letter.
Note that for multi-word terms you'll have to set Term consistency check in the Prefs to:
Please note also that this setting causes that you'll have to add any inflected forms manually, e.g. to a glossary equivalent of the memory: