We've addressed some issues, which should shave off at at least 700ms from initial response. Looking for further improvements across the board now.