Indexmap does reuse hashbrown. It consists of a hashtable containing `usize` indexes into a `Vec` which in turn contains the actual entries (keys and values), along with a cached hash for the key. In the end the overhead should only be that index and the hash.