Solution is this: Assign UUID to each async call, including retries. Or disallow retries. Sprinkle in some optimistic update if you want. Keep track of resu

Solution is this: Assign UUID to each async call, including retries. Or disallow retries. Sprinkle in some optimistic update if you want. Keep track of results, mirror lower level timeout (75 seconds for HTTP).