Some games do what you suggest, Daughter for Dessert (DFD) is a good example. The backgrounds are mostly pre-rendered, and the actor is set as an overlay, so for scenes with preset backgrounds, the rendering is minimal. In the bar case, these scenes are custom and each scene is different, so the original scene usually wouldn't be reused (but could be in some cases). This game doesn't work like DFD, so it has a lot more renders, and of course animations are just a ton of individual shots strung together, and this game has a lot of those especially with Abby kissing the MC so frequently. Because of all of those extra renders, that makes this game a lot more work than a typical VN would be.