The focus on _latency_ is a recent thing. VTE was previously focused on throughput and low CPU usage, which are conflicting goals.