Would be good to have __dask_tokenize__ methods added to Array and possibly Group classes. These methods are used to generate unique identifiers for different Dask objects. By default, they will be totally random if the objects in question don't define this method or register some function for making this determination with Dask. Having these methods will allow reuse of Zarr Arrays that are loaded repeatedly. Also this will be helpful for some applications like caching where deterministic hashes are needed.

ref: http://dask.pydata.org/en/latest/custom-collections.html#deterministic-hashing