Boy, this cetainly was a surprise. I've been watching rather intently the recent uptick in animosity between Merb and Rails. Merb started out great, promising a stripped-down Rails and thread-safety to boot. But in the end, Merb's claimed thread safety didn't actually pan out, and it lacked the polish that Rails has when it comes to programmer-friendliness. Plus, with the addition of Rails 'Metal', it's become very easy to chuck action controller altogether for request processing and directly hook up a resource-intensive task.
So, more and more, Rails was eating into Merb territory, and not surprisingly so. It's the same story with products - "faster X", "cleaner Y" and "easier Z" are always doubtful propositions in the long run because those X,Y or Z will catch up to you, and your raison d'etre is no longer there. But not just that - I found Merb and Rails are very similar in their structure, and very complementary goals. So it only makes sense that they merge.
What I would think is more interesting, or worth learning as opposed to spending time over Merb is even more bare-bones frameworks like Sinatra. It has a uncanny resemblance to Web.py, and that is a good thing. It's not always that you start out to write an MVC web app. Recently things have been getting very interesting in the Ruby world, with Github being the center of all this spurt of innovation. Exciting times in Rubyland, as always.