I don't think redundancy has been the problem at all. We all knew about the overlap from the start and I agree with iliasfl's point that users should be able to expect very different types of answers when asking a question that's on-topic on multiple sites, depending on where they choose to ask. The root of the problem in my view is that too many questions like these are being treated as appropriate/acceptable content:
I'm not saying every one of these questions should clearly be closed or that any of them necessarily deserve downvotes, but questions like these at some point came to represent a major portion of the site's content. I used to come read through the new questions here every day (when I had more free time) and I definitely started to feel like the noise was overwhelming the signal. That's really bad for a site that has yet to establish itself; it's a strong disincentive for attracting and retaining expert users at a time when building a strong user base is most critical.
It's not the naiveté of these questions that makes them bad content – beginners are perfectly capable of asking thought-provoking, well-researched, non-trivial questions. For that matter, even a trivial question can make for good content, provided it's well-defined and difficult to answer through secondary research (or the answer is simply not available online).
The main problem with these questions is that they don't fit the format of the network, which isn't something to be taken lightly. The Q&A format of Stack Exchange is a purposeful and central part of its design that has evolved over the years to better support that "repository of useful and accurate answers" mission. We discourage questions like the above – primarily by closing them – because they hurt that mission. To quote from Jeff's blog entry, Optimizing For Pearls, Not Sand:
It’s true that you can’t have Q&A without questions, but having the
wrong sorts of questions is far more dangerous. The fastest way to
kill any Q&A site is to flood it with low-quality questions....
Now, in our case it's not that we're being flooded with terrible, awful, no-good, very bad questions. These aren't questions nobody should ever ask, anywhere; they're questions that are of practically no interest to the community of experts that we need to attract in order to generate other high-quality content that in turn attracts better questions and builds and retains momentum for a young site.
This is the major flaw with the position that some have held throughout this beta that since we lack content, we should be less choosy about what content is acceptable. Actually, I think the opposite is true in a beta – the less content we have, the less signal we have, and the more important it is to weed out the noise. It's only by being choosy and scoping the site with the ultimate goals in mind that we send the message to the expert community: Stick around because this place is growing into something worthwhile.
This is closely related to the conclusion of the aforementioned blog post:
Answers are the real unit of work in any Q&A system. Therefore, the
only logical thing to do is to maximize the happiness and enjoyment of
answerers. If this means aggressively downvoting or closing unworthy
and uninteresting questions, so be it. Without a community of people
willing to answer questions, it really doesn’t matter if there are
questions at all, does it?
As to whether the site is "dying"... I think it would be a mistake to think that any beta is easy, if only we have the right policies or cast the right votes. But it may be that DataScience.SE is a particularly hard site to build, at least at the present time. The field itself is very interdisciplinary and very young; how many people really think of themselves as data science experts? Or think of their problems specifically as data science problems? I very much believe this site can and should exist, but the evidence seems to indicate that it's not practical in this incarnation at this time.