I'm interested to see what's said for "poor cache-ability". It's an area where the assumption of "more caching is better" can be wrong. I've seen real-world A/B test results that indicate client-side caching can often be slower than re-fetching a resource from the network. Typically, this is true for small resources being loaded from the disk cache (i.e. your home page on a new browser session). It's counter intuitiv…
Until some guy from New Zealand that doesn't have an edge in his country comes along. Don't actually apply this optimisation just because it's faster for you. It's probably slower for a lot of the rest of the world.
And who cares about the "guy from New Zealand"? You optimise to the markets/areas you target.