We currently do check if results for a training/inference jobs are available and give an error notification if that is not the case.
However, this is only possible if we can extract the flow-id for the flow that created those results. That is not possible if a training job fails and the inference child flow was never scheduled.