+1 for this fine explanation. However I'm well aware of how this works and my actual question was really what part of the problem in achieving this ideal scheme

+1 for this fine explanation. However I'm well aware of how this works and my actual question was really what part of the problem in achieving this ideal scheme Superpowered is solving. Or are they just targetting <10mSec latency? Are they getting rid of intermediate layers? Or are they just lowering buffersize? Or both?