Depending on what OS you are using, you can search a directory hierarchy for PDFs, then Ctrl/Cmd+Shift+drag-drop all found items into Zotero, which would add all PDFs as links (Zotero will freeze for a considerable amount of time if you try to attach 4000 pdfs, like... possibly 30 minutes. You can disable PDF indexing during import to alleviate this if it's an issue) You can then select all PDFs and retrieve metadata for all of them at the same time.

You mean that I can directly drop the directory?

No, you need to drag-drop the actual files, not the directory.

I do not mind to waint as long it does not crash ... I think that indexing files is important for the later search isn't?

I didn't mean to suggest permanently disabling indexing, which is important for searching contents of files, but I would suggest disabling during the import and then re-enabling once everything is done and let it go overnight. You can disable it in Preferences -> Search by setting Maximum pages per file to 0.

Do test this procedure with a handful of PDFs to make sure everything is working for you as you want.

I think that there is a limit to select all PDF to retrieve metada. Up to 80 it gives me back the option (right button of mouse), more not.