For me, I put them in directories that have to be there, because the underlying code doesn't create the directory, and without it, it fails.

Another example is where you want an empty directory mounted in Docker. If the directory is not there it is created with root permissions and then I can't even look into it.