I think I reported essentially the same thing not too long ago: Picard can't match AcoustID of a track, but the AcoustID is added. In that thread @Zas asked @lukz to look into possible replication issues, but there was no reply. Maybe he’s not reading this forum?