Hey Yoann Rodière, it seems that Analyzer.normalize method does not use the filterFactory field of the instance.
Thus no filter will be applied on the text.
I made this project to reproduce our case: https://github.com/fax4ever/lucene-play.
Therefore I think that we cannot:

simply use Analyzer.normalize instead of our own utils in org.hibernate.search.backend.lucene.types.codec.impl.LuceneStringFieldCodec#normalize.

Can we?