Imagine a workflow:

- clone a repo

- run a command

- run another command

It runs several times daily. Maybe it's CI or something.

Now suppose you want to cache the filesystem state for each command so that they can be rerun in a debug scenario where you'd expect them to behave the same as they did the first time because they have the same filesystem. (Having then recreated the bug, you could then start making changes towards a fix).

You either end up with many many copies of that repo, or you use something like this to only store the unique files and instead have many many indices into that store.