Given that the request for a pull to refresh API is a subpoint to "Scrolling performance", I am guessing they found that their implementation of pull to refresh had all the drawbacks they listed under "QoI issues".

From the post, "[Scrolling performance] is one of [their] most important issues."