My main incentive in treating files as entities is to allow you to reference them in text and connect them to multiple entities. This is how fibery fundamentally works so I feel it is not stretch to apply it to files. Having files (at least file names) searchable is just a nice bonus 
That would be a good option. I also think maybe allowing an integration with major cloud storage platforms so that files are actually stored elsewhere could also be a good short-term solution. Given that currently you cannot backup your files out of fibery or really do anything interesting like sync or edit, I think going in this direction in the short term might give users what they need without a lot of investment in the storage stack. Fibery can always add that service to the platform in the future.