This is something I have been asking myself lately. Sorry if I'm getting too much out of scope.
Whether an architect, product designer, UX designer, etc. a designer should provide an obvious and easy-to-use way of interacting with the object created, understanding its context, who the end-user will be, etc. This can be applied from an urban to a pencil scale.
In the case of products, its main functionality should be obvious ("Form follows function"). Once the functionality is known the way the user can make use of the product should be made in an intuitive way.
The main UI functionality should be enough for the user to manage the product in an effective and comfortable way.
Secondary functionalities let the "advanced user":
There should be a balance between main and secondary tasks, and there is where a good design shines over bad design.
In the physical world there is always a minimum of knowledge the user is assumed to have (such as that a button is meant to be pressed). This is the reason why digital UI tries to emulate (with less or more detail) the physical world (skeuomorphism). In the digital world it becomes more tricky as it is quite new and paradigms are not as established as we designers might think.
Physical products usually come with a book of instructions explaining how to use the product. Both main and secondary functionalities can be found there.
Now when we enter the digital world this is not exactly the case. Google Material design guidelines has a great page on Feature discovery which, to me, resemble an UI and UX instruction manual.
Your question:
Can we rely on users knowing how to use numeric steppers? Is it common for users to experience this kind of difficulties with the stepper?
Commonly secondary features become hidden and only accessible by users luck, or from mouth to mouth.
A designer should balance the users previous experience. He should create products which don't make the user think. He should give the user "feature discovery tips" to help him accomplish the tasks in the way they are meant to be done.
In my opinion, and as you experienced yourself, we can not rely on users knowing how to use numeric steppers. User research helps to understand how "hidden" a functionality is. And in such case a good design should be accompanied by discovery tips.