How does attribute lookup work on a Python object?

Question presented to candidate: "When you write obj.name in Python, where does the value actually come from? Walk me through the order Python checks things in, and explain why an instance attribute can hide a method but cannot hide a property."

What a strong answer should cover:

• obj.name calls type(obj).__getattribute__(obj, "name"); for ordinary classes that is object.__getattribute__, and __getattr__ is called only if it raises AttributeError.
• The precedence: a data descriptor found on the class (defines __set__ or __delete__, such as property) wins; then the instance __dict__; then a non-data descriptor or plain class attribute found by walking type(obj).__mro__; then __getattr__; else AttributeError.
• Why methods can be shadowed: a function is a non-data descriptor (only __get__), so an instance-dict entry with the same name wins; property is a data descriptor, so it does not.
• Assignment is different from reading: obj.x = v goes to a data descriptor's __set__ if one exists, otherwise straight into obj.__dict__ (which is why self.count += 1 on a class attribute creates an instance copy).
• Implicit special-method lookups (len(obj), a + b) go to the type and skip the instance dict and __getattr__; __slots__ replaces the instance dict with descriptors on the class.

Clarifying questions expected:

• Do you want normal instance lookup only, or also lookup on a class object (which uses type.__getattribute__ and the metaclass)?

Code / implementation expected: Optional — a small class showing a data descriptor beating the instance dict and the instance dict beating a method.