Isn’t voting no on possible good edits a bad voting practice, though? Perhaps you could try to ask the editor for more information first? I am aware that you are reviewing every merge (and I’m very grateful for that!). Perhaps after https://tickets.metabrainz.org/browse/MBS-10299 is implemented it could be manageable this way?
Additionally, even though the following link is not a guideline, I personally try to base my current voting tendencies on the following article: Editing: Making MusicBrainz better – MetaBrainz Blog , although sometimes instead of politely asking the editor to cancel the edit first I do cast a no vote if their edit makes the data worse, however I always accompany it with an edit note explaining why. Also I can’t approve and fix stuff myself cause I’m not an auto-editor (doesn’t matter for merges in this case anyway).