We currently autogenerate models based on the order that the Camunda event sources were supplied. We should instead base this on the order of known correlations between the event sources themselves. This may require some means of sampling. For a first iteration, it is probably ok to have linear connections, with more complexity, such as parallel sources, coming later.
- The order of the event source models in a multiple source generated model is based on known traces between the sources rather than just sequential