The rest was just demonstrating that it works, and produces pretty reliable random output. In order to do that, they used a standard tool that tests randomness. But it's nice to make sure your randomness test tool is actually good; so using it on a known-bad RNG, and seeing that it does fail, is a good way to demonstrate that.
Apparently a lot of android devices don't have one, and don't have time to collect pseudo random data for their software rng to make secure certificates when you boot it for the first time, which posses a security risk for android devices.
That's how I read the article in relation to recent news.
I'm guessing not everyone on HN is a huge fan of “Strictly Ballroom”, as the title is a key quote for dance movie otaku. It was also a weak joke about testing for random numbers; I'm imagining a little subroutine in a constant state of surprise when it encounters the next random number.
While I'm not sure if I implemented it correctly, RANDU isn't some ‘lousy homegrown software RNG’. It was the standard RNG for IBM mainframes in the late 1960s.