"Why do they need a clock at all?"
Yah know, I have a particular weakness for multiple serial links for high throughput. Sure, it takes a bit more logic to take the packets apart and putting them back together again, but transistors are cheap...
Two reasons I can think of is that to get an equivalent data rate, a single serial port is going to draw more power and be more difficult to route on a pc board than a wider bus. Not real sure it's that much different, though.