That is certainly a big usability oversight (especially with large PDFs), I remember Chrome used to do that with images as well.

However, if those PDFs are requested via HTTP, unless the bank gives appropriate caching headers I think Chrome is technically correct in re-requesting them: they might have changed in the meantime. (I assume it does the same with webpages when you save them.)

The fact that the bank gives you a URL you can only HTTP GET once does sound like very bad implementation on their behalf. Perhaps it's a cookie issue, or even a bug in Chrome itself?