Anna, ok got it. I’ll ponder our model a bit. 
I guess where I’m struggling is that in our current model our desired flow makes incremental changes to data in each step. If that flow then fails in step 5 out of 8 for example (after a few retries of step 5), that piece of data is left in a state of result after step 4. We may need to evaluate the issue and when resolved, then pick up from step 5 again for that piece of data.
One approach could be to ensure that steps 1-4 (as well as other steps) are smart enough to check for the state of the step and not re-process the data if it has already been done. This way we could just rerun the entire flow with the parameters as needed and steps 1-4 should breeze through if the works already been done.
Alternatively I could pass data from one step to another in the flow, but in this case I’m weary about that approach since (a) the data is quite large and (b) the processing times for steps 1-4 can be long so I’d rather not replicate work that’s already been done.
I’ll try to read up on some more orchestration approaches to see if I can come up with any ideas. Seems like my first approach would be the best though.