When is Gemini’s long context more useful than uploading a document?

I’m comparing ways to work with a large set of notes or reports in Google Gemini. If the goal is to ask questions across several documents, when is Gemini’s long-context window useful, and when is it better to upload files or split the material into smaller sections?

I’m curious about practical limits too: does the choice affect how easy it is to check where an answer came from?