Had some time to internalize your feedback now - So to put it in infrastructure terms, no specific considerations, other than the regular ones dictated by the traffic flow required to the cluster, regular network policies, wherever they are enforced (centrally or host based, or both) are sufficient based on the nature of the service to control it’s exposure.
As for throttling, very interesting point. And a further question! As mentioned I’m way out of my knowledge zone as to what is the application of this solution in terms of use cases. Have you seen any use case generating enough data to be considered for throttling? This is also a consideration for perimeter placing of this - based on the service consumption model - IMHO.