The Google Cloud standard for async operations via the google.longrunning.Operation service: https://github.com/googleapis/googleapis/blob/master/google/.... This should be usable either via REST/JSON or gRPC.

It's designed to be generic for any kind of async operation, and to be "mixed in" with your APIs. There are utilities for waiting for an operation to finish (via polling); in theory it should be possible to use some type of server-push to avoid polling, but I'm not sure if anybody's doing this.

(I'm an Alphabet employee who's working on APIs delivered via gRPC / REST/JSON.)