This is because of the low percentage of devs in the survey who have used Jest. I can't begin to describe the joy of finally finding a JS test utility that just works and doesn't require a giant configuration file.
If you aren't using Jest, make the switch now! It's likely a lot of your existing tests will work with Jest.
Most of that is just find-and-replace kinds of changes. But we dropped 3 dependencies to one and sped everything up in the process. It's just a fantastic tool! I finally feel like I've got a good Rspec equivalent in the JS world.
I don't know what the general feeling about e2e testing in the JS community, but in my last job we had to write, fix and maintain hundreds of e2e tests and it was a nightmare.
Our stack was Angular 1 and the de facto Angular e2e tool, Protractor. We wrote tons of helpers, like waitToBeVisible & waitToBeHidden, most of them copy/pasted from StackOverflow because every users of Angular and Protractor have the same basic troubles and use the same workarounds, and we had to use those helpers in every test. Also I can't imagine the number of hours spent to mock our Angular services + MySQL/Redshift data, to spy the functions who manipulate url/cookies/localStorage/etc and to write some "page objects" to make the tests readable.
In my opinion: The benefits of the e2e tests are too small to justify the crazy amount of time spent to make the new tests work and maintain the existing ones. Most of the time, the tests break because someone has changed something in the DOM or renamed a CSS class or modified an external API and forgot to update the tests/mocks. But they almost never break for a regression... so what's the point?
Or otherwise you need a dedicated team with 1 or 2 engineers (for a small/medium company) working full time on the e2e.
(PS: I used only Protractor with Angular 1. I haven't tested the other tools like Nightwatch.js, and I don't know what they have done for Angular 2 (Protractor 2?). Now I develop in React and just do unit testing & dog fooding).
I think the benefits come when you're really making use of reusable components. I have about 20 views that all make use of the same Redux container component, and e2e tests would definitely help me sleep better at night.
https://facebook.github.io/jest/blog/2016/10/03/jest-16.html
If your test is fully CPU bound and limited to one core, it probably won't be faster with Jest. If you split it up across multiple files, it will parallelize tests.
A lot of people in the frontend world equates unit testing with Selenium tests (aka: integration tests) and get really ruffled/confused by "pure" unit tests.
Now Jest changed the defaults to what most people expect, which will confuse them less, I suppose.