I never understood why hybrid vector/bitmap editors didn't gain more acceptance: it seems a natural way to work. This one-or-the-other mentality may suit developers of the software, but it's not natural for users.

I like to be able to go back and edit my shapes, yet change the contents in ways that only a bitmap editor would be able to. Storing a history of the transformations I did and re-applying when I'd resized/changed the shape seemed sensible - as well as allowing me to go back in history and remove some.

As a self-taught designer, watching professional Photoshop users for the first time was mind-boggling: they make items and then throw them away and re-make them if they need a different size - or crop parts off, rather than masking. It's all destructive editing! I learned a lot about layers and could see the benefits of them, but still...

If I remember correctly Deneba Canvas supported the hybrid Vector/Bitmap approach, but never gained enough traction to be a viable product.

I notice this about Photoshop users myself but I also notice the more time you spend using it, the biggest efficiency gains you make to your workflow come from using vector shapes and non-destructive editing (various types of masks, adjustment layer, layers with blending options). There are a few things where destruction is required but often I stash an old copy of the layer unless I'm dead set on the new version, but this can be rare. You also learn as time goes on how annoying a PSD with rasterised elements that didn't need to be rasterised are!

You should always be able to do non-destructive editing even when pretending to work destructively, by keeping around the original sources (which you keep around anyway in most cases) and the graph of operations applied (which isn't much data). Or rather, the software should be able to do that. I'm surprised that there don't seem to be any applications doing that.