My (somewhat rambling) 2cents, since I started working on improving this via OCaml Labs.
- Poor tooling: OCaml (and ML) is taught at a number of institutions so undergrads do have experience of it, yet few places seem to use it in production. The stories I heard of how they had to manage their systems would have put me off from ever wanting to learn it. OPAM, and the new tooling around it has changed all that.
- Fragmented community (outside academia): Of those who've used OCaml, there are those who love the language but they may end up feeling a bit isolated as they're unaware of others who can help/support them. They either soldier on alone or move on to something else. This shouldn't be surprising if you believe the 1/9/90 rule -- 1% make/do stuff, 9% use/participate and 90% just consume. I'm suggesting that the quiet 90% aren't really aware of each-other's existence and don't progress into the 9%. The work around refreshing ocaml.org, CUFP, more blogging etc is helping this.
- Missing 'killer app(s)': Using a language for it's own sake isn't very compelling. There need to be tools or products that people (devs) find impressive or cool that have been written using OCaml, where it was clearly the right tool for the job. This drives interest in the language with a purpose and encourages people to make/contribute things. I believe Mirage OS, Hack and recently Flow have helped this (as well as the effort of those projects to actively disseminate their work - see above comment).
With the above in mind, I had a couple of thoughts in the early days of OCaml Labs (I'm sure these weren't original).
Thought 1: Improve on all the above and existing users who've quietly appreciated the language will feel more confident in revealing themselves to support a growing ecosystem.
Thought 2: If thought 1 is true then some of those, who've come out to support the ecosystem can feel more confident in advocating for it and draw in new users. Combined with the above, this sets up a virtuous cycle.
As a result, I'd expect to see more stories on HN, Reddit, Twitter elsewhere that refer to OCaml, either as a language or because of tools built with it. In other words, this shouldn't be a fad but merely revealing another option to solve certain kinds of problems.
Your work is showing the results and its really inspiring.
I believe a major problem with OCaml's adoption always was that a large chunk of the community and documentation is in French.
Maybe the language is finally hitting critical mass? Maybe lots of people are hitting the same shaped problems at this stage in the evolution of our cloud platforms? Maybe it's just dumb luck?
For your specific things: Rust is probably not mature enough yet but worth keeping an eye on. Go makes me cry. Scala does a lot of the same things OCaml does as well or better, at least if you "stick to the good parts" (there's a ton of extraneous crap in Scala); if you're in an environment where the JVM is acceptable then TBH you won't gain ever so much from having both in your toolbox.
Haskell is the tough one. My take is that purity is a very heavy burden, and even the most advanced language in the world can only go so far to alleviating that. But the power it grants you can be huge - or can not. If you find purity worthwhile, go with Haskell. But if you just want the functional power boost and don't want to immediately expend most of it on carefully tracking your effects, OCaml is a perfectly good choice; at the least, it's an incremental improvement in the Haskelly direction, without having to wear the hair shirt right away. I'd say much the same for Scala.