This is not an alternative way of using selenium. It's a way to avoid using selenium. The advantage over selenium is that tests and application code are run in the same process, so they tend to be faster, less brittle and easier to debug. The downside is that they are not cross-browser.
There are also other ways of using Selenium, though, that I had mentally disregarded and ignored as probably not useful.
You can also get this behavior (basically single thread of execution) with cucumber-rails and Selenium (or probably any driver other than Selenium, for that matter). I'm not saying the application code is literally run in the same thread or process as the test executor, but in effect they do. Your Ruby process is in charge of setting up the HTTP server, as well as driving whatever driver (Selenium, Phantom, ...) you're using for Cucumber.
So serendipitously, with (MRI) Ruby being a single-threaded beast, you do still get that behavior and quality of being able to catch errors that pop up on either side, making it much easier to debug misbehaving server or tests.
