Users of a microwave currently expect extremely poor temperature control, and so don't even think about temperature because you get what you get. An interface which lets me heat something to exactly "drinkable but hot" is a revolutionary change in how I use the appliance though.
Same story with warming food: sure, you want it warm all the way through - if you actually know the surface temperature though, then you can model that - food that is cold inside is going to have different surface temperature decay rates to food which is warm inside, and some modeling of the observed thermodynamics means the appliance can know this.
So now we'd really be talking: "warm to safe temperature all the way through" and we can give a probability / % complete read out of how close we think we are to it. No more guessing, no more user interaction, also no burning yourself because some parts went superheated.
For cooking people will either use recipe from a book provided with microwave and thus they need just a selection of a program, or they will use more advanced microwave with oven and will want to set temperature of the oven to something like 200 degrees for 40 minutes. In this case sensor would be useful, but it's already not a microwave story.