Make runAutomatically more resilient

We need to delay running the result or else we might end up clearing the model
whilst in the middle of processing a model insertion