We currently only attempt to autogenerate a process based on external events if we can find at least one start and end event, which is where an event type features in sequence counts as a source event but not as a target event (and vice versa). Given that a single incorrect trace could eliminate an event type from being considered a candidate for being a start/end event, we should find a more flexible way of determining a start/end event where at least one can't be found in the current way. An example would be to say that an event that features as a source event and only infrequently as a target event is a good candidate.
- We have a secondary means of finding potential start events for external event source autogeneration
- It is acceptable if this results in not all external event types being part of the resultant model
- This might prevent looped models from working at the expense of being able to create a model at all