How to handle distributions no longer available

What should we do if:

1.  a previous run of pipeline-void has processed a dataset, producing a VoID summary, 
2. the dataset and/or its distributions become unavailable
3. a subsequent pipeline-void run currently doesn’t write anything, so the old data is stuck in the knowledge graph.