OCR data being available for search depends on the data being indexed by the search engine, and I think the note-resource association also being indexed. Maybe some of this doesn't happen quickly enough by default and is indirectly triggered by sync.