Edward Robertson has a thought-provoking post on whether or not relying on Amazon's KDP Select is a viable long-term strategy.
It will probably not surprise you to know that I think that it's not, for many reasons, but of course I also think it's a totally reasonable short-term strategy. I'm a fan of experimentation, as long as people stay nimble and don't lock themselves in (and one of the nice things about KDP Select is that it is a short-term agreement--three months instead of the life of the copyright or something like that).
Diversifying is always difficult, which is why even in the realm of personal finance, there are people who think you shouldn't do it. When you diversify, you are almost always forgoing easy money now in favor of security later.
For example, when that long-suffering encyclopedia editor wanted me to work for him all the time, that encyclopedia was paying better than my other clients. I didn't have to prove myself to him or spend a lot of time sending out pitch letters for people to ignore--he was happy to give me more work than I could possibly do.
And yet, none of that changed the fact that that project was very suddenly shut down (another case of "we don't want to pay to print this"). Had I been solely dependent on it for income, I would have made a lot of money in the short term--and been deeply screwed, with no contacts and no clients, in the long term. The problem wouldn't have been simply one of income--it would have been the fact that I hadn't kept in touch with my other clients, who sensibly enough, would have found other freelancers to do their work.
In the world of finance, it's really easy to make a decision look good or bad by fiddling with the timeline: If you bought your house in 2004, by 2007 you looked like a genius, but by 2009, you looked like an asshole. It's the same thing with diversifying: For the first--oh, I can't remember, 12 months maybe?--I looked like an idiot for spending time on other, less lucrative projects. But two months later, I looked a whole lot smarter.
My point is that KDP Select is a tool that seems to work pretty well for some people. But it's not a perfect tool, and it doesn't work for every writer or for every project (and has never, even before the changes to Amazon's algorithms). Most important, there are other tools out there, especially when it comes to marketing, and many of these tools (garnering reviews, advertising campaigns, pushing paper books to stores) can be explored while a book is on KDP Select--it's not always a case of either/or.
If one day your KDP Select tool suddenly vanishes, you'll be happy you went to the trouble of figuring out what other tools work for you.