Those are all good things, but they aren't really the same reason. Parent describes JQuery to React as spaghetti code -> reasonable code, whereas React to ReasonML sounds like reasonable code -> reasonable code (with better tooling).
Yeah, I agree but would say reasonable code -> slightly more maintainable, reasonable code. Sure, the low hanging fruit is gone, but it's a similar reason. What I'm getting at is that you can write spaghetti React pretty easily, and people do. One of the core promises of using an ML to do react is that its harder to mess with state all over the app and wind up with a big bowl of spaghetti.
JS code can be a pain to maintain (in the membrane... joke, sorry, I'll stop). And FB knows this. That's why they developed ReasonML: great JS interop, but much better maintenance story (than both JS and TS) because of the stronger typing discipline.
On a large code base, increased maintainability can be a huge commercial advantage.
