> I fixed this by defining an inode(string) function which hashed a string to get the inode number, and using the tree ID / blob ID as the string to hash.

The tree/blob ID is already a hash though. You don't need to hash it again. Just use the first 8 bytes of the tree ID as the inode.