The surveys in all of these years were open for different amounts of time and advertised on different platforms at different frequencies. I think primarily they show the limit on the number of people willing to take an online survey.

The big jump from 2014 to 2015 coincides with the emergence of the Clojure Slack community. As a real-time chat platform it tends to collect more people and make it easier to find people in the community that are available right now to answer a survey.

I think it is very hard to conclude from this data alone anything useful about the number of users of Clojure. I watch metrics on a number of platforms (some public, some private) and my sense right now is that there at least a small spike happening over the last few months, but again it's very difficult to conclude anything with certainty.

However, judging by StackOverflow Trends [0], it's even dropping.

[0] https://insights.stackoverflow.com/trends?tags=clojure

In the current trend of Node-based backends, the combo of "share code on frontend & backend" and "use your existing Java backend libraries" is no longer so distinctive, because there are lots of npm libs available now for the backend. And there's the microservice trend, which also makes it easy to integrate JVM components into an otherwise Node based backend.

Clojure(Script) is still a much better language than JS of course, but the other factors offer less additional boost now than in 2015.

There seems to be a backswing of the pendulum now against microservices. Maybe we will get one for Node backends too?