>Also if Batch C is being sold or something similar (to batches), a better idea would be (IMO) to have the product owner comment on the review, saying we have fixed this in batch 2 or batch B or something
That is a good idea, but that still does not fix the aggregate data. When someone searches for "Coffee", the aggregate number of reviews is still considering that low-star review. So the "Seattle Coffee Beans" overall rating for Batch C could be a 4.5 / 5, but for all batches is a 2.5 / 5.
The easiest way to fix this is to remove old, out-of-date reviews. This should apply to positive reviews too, but I have no idea if Amazon is doing this.
Another approach is what Steam did - have a long-term rating and a "recent" rating. See "No Man's Sky" as an example [1] - a product that started poor, but improved over time. This helps prevent product brigading and gives vendors an opportunity to bounce back from negative reviews.
[1] https://store.steampowered.com/app/275850/No_Mans_Sky/