POC on how verification function would work. Update function should have something similar.

realis.go

@@ -672,7 +672,23 @@ func (r *realisClient) CreateJob(auroraJob Job) (*aurora.Response, error) {
 		func() (*aurora.Response, error) {
 			return r.client.CreateJob(context.TODO(), auroraJob.JobConfig())
 		},
-		nil,
+		func() (*aurora.Response, bool) {
+			getTaskResp, err := r.client.GetTasksStatus(
+				context.TODO(),
+				&aurora.TaskQuery{JobKeys: []*aurora.JobKey{auroraJob.JobKey()}},
+			)
+
+			if err != nil {
+				return nil, false
+			}
+
+			tasks := response.ScheduleStatusResult(getTaskResp).GetTasks()
+			if len(tasks) != int(auroraJob.GetInstanceCount()) {
+				return nil, false
+			}
+
+			return nil, true
+		},
 	)
 
 	if retryErr != nil {

retry.go

@@ -197,9 +197,12 @@ func (r *realisClient) thriftCallWithRetries(
 						// Allow caller to provide a function which checks if the original call was successful before
 						// it timed out.
 						if verifyOnTimeout != nil {
-							resp, ok := verifyOnTimeout()
+							verifyResp, ok := verifyOnTimeout()
 							if ok {
-								return resp, nil
+								r.logger.debugPrint("verified that the call went through successfully")
+								// Response might be different than the original.
+								return verifyResp, nil
+
 							}
 						}
 					}