There is a bit of survivorship bias there, I've seen a lot of Java projects fail, but the blame is only partly on the language, really its the fact that there are so many developers, and the pareto distribution of good to bad java programmers. That and business do not know what the f*ck they are doing. I've seen really good java devs at work, the software they produce is amazing, but the code and structure is incomprehensible, they are masters and I am a mere mortal. Clojure on the other-hand is simpler, higher level, more expressive, but you can still shoot your foot, just like java.

you cannot measure anything at all about failure rates by looking at successes

But are the rates for Clojure failures more or less than Java?

It is unlikely you will find apples to apples comparison.

What if Clojure projects are more risky by selection bias?

Maybe not at unusual rates, but I wouldn’t be surprised if the rates are high.

When you measure the failure rates of Java projects against the failure rates of Clojure projects, you’re not measuring the failure rates of a language, but rather the general failure rate of software in industry vs the failure rate for a (likely) small, motivated team.

Higher than average? That's very much a [citation needed] moment if I've ever seen one.