Currently several failures are not even visible as workflow failures: only looking in the log reveals the issue: The problem here is that playground-snapshot return value is used to describe if a new version of metadata was generated... This mechanism should be changed so that python execution failures do not look similar
The issue should be filed to to the metadata repository.