I'm not sure if this is even possible in, but would doesn't hurt to ask I reckon.
Would it be possible to have the second parameter for Merge and MergeDeep have the keys of the type you're passing into it be provided as possible overrides/restrictions?

type Foo = {
    foo: string;
    bar: {
        baz:  number
    }
};

type MergedType = Merge<Foo, { baz: string } >; // <-- `baz` would cause an error because the key does not exist on `Foo`
type MergedType = MergeDeep<Foo, { bar: { foo: number } } >; // <-- `bar.foo` would cause an error because the key `foo` does not exist on `Foo['bar']`

We've built an automated type harvesting system that collects a lot of type data to be used by the frontend and sometimes we have to correct those types deeply

At the moment Merge works as expected, but if any of the types for the response data changes we don't know that because Merge does not care whether or not the second argument matches the structure of the first.

Also apologies if this already exists, there's so many types in this library I can't wrap my head around all of them.