Maybe scripted updates in the elasticsearch output might work for you?
This would be in a second LS pipeline that reads the CSV file. Be aware that the document you are trying to update must be in ES already. You may need to use the elasticsearch filter to add the sequence_id of the first doc matching against order_id. Alternatively you could try to add an array of sequence_id of all matching docs by order_id and then split the array into multiple docs.