We don't use underscores in our JSON format anywhere else to my knowledge. We should fix these to use hyphens, as SOLR does already.
Picard was referencing the underscored names, but that's already been fixed and released:
PICARD-1471. In fact, Picard was bugged because it was trying to use the underscored names on a SOLR response (it doesn't perform any artist lookups), so it's an example of the differing property names causing unintended bugs.