When we did need to propagate changes back, we used a service by another team, if that team used queues, just a direct post, error to log. If they didn't, we pu

When we did need to propagate changes back, we used a service by another team, if that team used queues, just a direct post, error to log. If they didn't, we put up a queue, and added it to our queue processor and tried their service for 3 days, logging each time to eventual dead letter our ops team could retry.