This is because we are loading whole files into memory when processing them. We have a solution for streaming file contents into memory incrementally instead.

That said, since multi GB files are not particularly common place, we will put this in the backlog for now until it becomes a problem for customers.