I want jobs to not block the next stage, yet stages block next stage until all stage jobs complete, even if allow_failure is true.
This isn't ideal as it means stages take longer to complete than they have to.