It seems though, that ZORA implementation was broken in the version 6.0. So the *_zora_* tests should have failed.
I've tried some other ZORA computations with version 6.0, and they gave me strange results, few tens kcal/mol off. DK was OK, for the same test molecules.