I agree with most of what you’ve written, too often MVP’s are created as way of doing the least possible work to get to launch, when they should be about testing and verifying the key assumptions and unique value proposition for the product. Most of this can be done using prototypes and guerilla testing with small user groups rather than putting something half-baked out into the world.
In your example of a bakery, the MVP is not opening a shop and selling raw dough, but creating beautiful doughnuts and selling them from the back of a van. Then you’re testing whether there is a demand for your product.
To me, an MVP is testing your Maximal Value Proposition. The MVP is ‘we sell beautiful doughnuts’, not ‘we have a bakery’