...in order to avoid to have a MatchAllDocsQuery on Solr-side in AmetysQueryWeight when computing ids, because it generates an HashMap which takes huge space (300Mo if 2.5M docs for instance).

Try to have DocumentTypeQuery as much as possible