There are no absolutes in UX design.

Saying something is always bad UX because it has the potential to be so, doesn’t mean it’s true. In the same vein, if a user has to exert cognitive load, it’s not always a bad thing.

Based off of my tests and time with users, that cognitive load can result in learned behavior and can actually lead to users wanting to stay in the app because they are familiar with it. There’s an intrinsic quality to being an ‘authority’ with something, even something as trivial as a phone app. This notion of something not being obvious in function can apply to any modern interface. The iPhone home button, for example, has no icon, label, and even when it did have an icon, it was very ambiguous. Did that mean it was bad UX? I’d argue no.

Also, just because something has a label, are we to expect that everyone will interpret the word in the same way? ‘Compose’ in gmail applies to messages, but ‘Compose’ in garage band could apply to starting a new musical project. Isn’t that equally as confusing? Pencils are synonymous with both writing and email and editing a photo; are we to say that it should not be use just because it’s the same icon?

Icons may need labels in some situations, but in others they do not. It’s all based off of your specific users and their specific needs. Testing is the only way to find out if your users can discover and understand your UIs meaning. Being hyperbolic is not helpful, after all, some people may like mystery meat.