How do Unpack and TypedDict enable type-safe **kwargs?

Question presented to candidate: "Say a function takes a bunch of optional settings through **kwargs — timeout, retries, verbose. How would you type that so a checker catches a misspelled keyword or a wrong value type? What do TypedDict and Unpack each contribute, and what happens at runtime?"

What a strong answer should cover:

• The old form **kwargs: int types every value the same; it cannot give each keyword its own name and type, and inside the function kwargs is dict[str, int].
• **kwargs: Unpack[SomeTypedDict] (PEP 692, Python 3.12; typing_extensions before that, and typing.Unpack itself exists from 3.11) tells the checker the function accepts exactly those keyword names, each with its own type.
• total=False, Required[...] and NotRequired[...] (PEP 655, 3.11) decide which keywords are mandatory; required keys become required keyword arguments, the rest optional.
• Checkers report unknown names, wrong types, missing required keys and a named parameter that overlaps a TypedDict key; inside the function kwargs has the TypedDict type.
• Nothing is checked at runtime: the annotation is metadata, kwargs is a plain dict, and a typo still runs. Validate at the boundary if the input is untrusted.
• Trade-off against explicit keyword-only parameters: Unpack shines when the same option set is shared by several functions or wrappers that forward **kwargs.

Clarifying questions expected:

• Which Python version and type checker does the codebase use (3.12+ typing.Unpack support, or typing_extensions)?
• Do the options come from Python callers only, or also from untrusted data that needs runtime validation?

Code / implementation expected: Yes — a TypedDict, a function using **kwargs: Unpack[...], and the checker errors it produces.