As a native Japanese and sort-of English speaker, I need to pause flush audio pipeline to switch between languages. Konnichiwa spoken in the middle of an English sentence and こんにちは in native Japanese, are separate expressions and have to be pronounced differently. From my experience there likely isn't a single unified language model inside a human head, and if so it's not a surprise to me that one cannot be effortlessly made for computers.