Thanks for reading. Do not use a SAGA pattern in practice, can't tell much about it. What I would do in case of failure is something similar. I would use a queue, for instance, AWS SQS. When some part of an update gets an exception the message is created and posted to the queue. Then some another service polls the message and tries to call the service that returned an exception with the information contained in the message. Here we can configure the number of retries and the backoff timeout and its type (linear or exponential). For cases when the flow heats max number of retries, we can implement the dead letter queue (DLQ) pattern.