It's not just local context. Reading a dense book is still more difficult than reading a less dense book, given a fairly similar amount of information and style in conveying that information. Larger codebases suffer the same problem you mention in a different way, and cargocults in most static languages tend to advocate very verbose writing styles.

Where this falls apart, the more verbose writing style hasn't been proven to convey more information or in a better way. That's an assumption still tossed around.

It is just a fair bit harder to figure out the types as program grows.

Not in my experience. Language engines are good enough to help most of the time,

And typed or untyped, you’re only ever reasoning about the types in the context you’re working in, not the entire program.

Can you actually determine things like that with an engine?

A Python function can be called from within or from outside of your codebase with different callsites passing different types.