[mindundermatter]

How is this done and why would you it?

Some of the best engineers in Nashville will actually ride the gain on the preamp while tracking the vocal. This is especially helpful with a really dynamic singer/song. That's one way to do it. Compression can suck the life out of a vocal track really fast. Going back and automating the track brings back some of that life and really helps the track 'breathe'. It's not that difficult to do, but it can be time consuming. The way I was taught is to do an initial pass of automation (which will put you in the ballpark), then go back and zoom in really close. Then you just look at the waveform (I know that's bad; you're supposed to be using your ears, but this is the ONE time I can think of where visualizing the audio actually comes in handy) and where there are peaks you nudge it down, where there are valleys you bump it up.

Maybe the guy was just exaggerating how he automates!

Nope. There are actually people out there who get paid to do just this kind of 'editing', simply because it is so painstaking.

The point of overactive vocal automation is to make a compressor work less hard.

That's exactly why they do it. It allows for less aggressive compressor settings and a more expressive (and therefore 'natural' as you put it) performance.