[from business perspective]
The news cycle, attention span, and ability for anyone anywhere to blog/twitter about topic and "steal" your traffic. Makes breaking news and reaching crit-mass attention now! Makes A/B unrealistic and not worth the investment. Even on "static" items like this researched article you see how they failed to be firstest with the mostest and lost the traffic.
[from journalistic perspective]
They A/B "test" at the journalist level. They hire several, the ones that write popular / prize winning articles are kept, the others get downsized.
We considered this with NewsTilt, but never got far enough along to actually try it. The aim wasn't actually A/B testing them, but to use machine learning to predict how well the article would do given certain criteria: title, what time of day it was released, etc, to allow journalists to tweak those criteria. I think that was actually PG's idea.
