The problem is that it kind of relies on a lossless layer 2 (flow control) which has its own set of problems in large scale networks. This is what things like t

The problem is that it kind of relies on a lossless layer 2 (flow control) which has its own set of problems in large scale networks. This is what things like this try to solve: https://cloud.google.com/blog/topics/systems/introducing-falcon-a-reliable-low-latency-hardware-transport https://cloud.google.com/blog/topics/systems/introducing-fal...