But that's great until it isn't.
You can design all the declarative things you want, but someone will eventually come up with a requirement that doesn't appear to fit a declarative structure, such as an event that fires 2 seconds after something else happens.
And so to resolve that you can construct state machines with transition states to get back to a declarative model, but those state machines can quickly become a bigger headache to maintain (yes, even with redux) than if you'd just been able to setTimeout from the component in the first place.
And you start with the best of intentions of re-usable components but find more things that ought to be owned by the component pushed up to the store and suddenly there isn't a neatly re-usable component but an empty shell of something that still needs the other half plumbed in manually every time.
I love the idea that you can abstract everything happening in an application to state and render reproducibly from that, but the reality is that it's really hard and time-expensive to do compared to jquery.
But that would be fine if the tooling was there. If React had a CLI like the AngularCLI it wouldn't be such a problem to add new actions, or new pieces of state, or new components.
But it doesn't, so each new component ends up adding a little more state into the state machine which ends up breaking types in far off places.
Now there's probably an assumption about how state is built up or how reducers are typed or how actions are defined that I'm missing some reason why this shouldn't be happening. But it is happening, it's the reality for the project I've been working on and it's a real headache.