Does anyone know a) how this works under the hood, i.e. are they using phantomjs to handle apps with long-standing connections, or b) how they measure latency: page load time, DOM render time, etc?

Got some promising initial results for my app but it's time to try some science...