There has been a request for butler to have a special dimension (maybe a UUID as value?) that would allow datasets to be stored in collections that are not directly related to a dataId.

For example, if multiple cutouts are extracted from a single coadd they can not currently be stored as independent cutouts but must be combined into a single python object and persisted with the dataId of the coadd. If an extra dimension was available that was effectively a unique label, then the cutouts could be stored by adding that dimension to the coadd dataId.