Browser.js doesn't enjoy 100% test coverage. It was only partly increased in T203818.
I've been exploring the NYC coverage tool and discovered that it can produce pretty nice HTML reports (by changing the npm coverage script to nyc --reporter=html npm -s run test:unit).The output of these reports can give us more insight into what is and isn't being tested.
Here's the output as of the latest patch https://people.wikimedia.org/~jdrewniak/coverage-report/. As you can see in Browser.js, the lines that aren't run relate to: the memoized function, iOS version, and getSingleton method.
Knowing this now, I think we can increase the code-coverage of the Browser.js tests.