You can use Google Analytics as a guide, but we'd expect our count to be lower than what GA will report, given that GA will not differentiate between views loaded from the XF cache rather than generated live etc.
The underlying rules are relatively complex, but as a general rule, we only count requests that connect to the database and render most of a page of HTML.
By no means am I suggesting the cloud sites are slow (they obviously aren't). But, genuine question. I tested a handful out of curiosity and they seem like they could be better? I don't know where each site's server is located but I used Vancouver, CA for each. My site's server is in Washington, USA for reference.
My site is tacoma3g. I do my best to optimize it and I'm thinking it did better because of the Font Awesome Manager add-on. But the comparison sites are using default/simpler styles. If anything, I thought @Brogan's would be a little better. But again, maybe it's just the server locations.
Yes pagespeed involves many factors. Default XF 2.x style has some things that slow it down page speed wise which I've managed to shave 0.3 to 1.0s off LCP for on my test XF 2 forums using XF 2.x default style or PixelExit's Flatawesome+ style.
But also helps to have CDN caching in place so that your site/forums are fast from every geographical location i.e. test server response time (TTFB) using https://tools.keycdn.com/performance and https://speedvitals.com/ttfb-test and you can clearly see how fast your site/forum is from cities around the world for that initial TTFB which factors into every other Google Core Web Vital metric like LCP.