This is the problem with Tailwind, it doesn't offer enough nor does it move faster than CSS to justify its existence. The only real use case for it seems to be solo developers that do not have the time or resources to set up their own baseline set of variables, or those that use it with a component library that's based on TW.
Personally doubt the better performance, I have yet to see a case where that is true unless they were comparing it unevenly to some behemoth like Styled-Components. When you compare it to SASS the performance difference ceases to exist. Same with "improves the dev exp", I don't see how this is the case at all especially when you have to install an additional extension to get full proper support in your IDE of choice. You are still working with variables, if you are on a tiny team/solo, it makes sense to use Tailwind defaults but on a large team where your design team should be creating their own? What use does Tailwind have then when the utility classes are just going to be overwritten? You're still creating the components yourself.
.button { background-color: blue; color: white; padding: 0.5rem 0.75rem; border-radius: 0.25rem; }
Then the browser will need to download less bytes:
<button class="button">Click</button>
Compared to:
<button class="bg-blue text-white px-3 py-2 rounded-sm">Click</button>
And this does make a difference when your entire layout is done with Tailwind. Consider that the HTML, unlike the CSS, might typically not be cached, and that the server is now also charged with emitting higher loads of data. Consider also that all these utility classes are all loaded in the CSS on the front page even if some or most of them is used on everything except the front page. The performance win is not as unquestionable as Tailwind would like to suggest.
We already get massive benefits from it that I wouldn't get as a solo dev. If someone else writes a bunch of new components that make up a new menu and I need to grok them quickly and then start working on them quickly, Tailwind brings the time I need to spend understanding the styling down to nearly zero. It's a massive timesaver for teams of any size.
Yes, you can get some of the benefits with other systems like CSS variables. But not nearly to the same degree, and you'll also need to organize a bunch of CSS/SCSS files besides you main code.
With Tailwind we have something like five small CSS files across a large monorepo that contains several apps.
Imagine being a computer programmer and not thinking nor doing math in your head.
Good thing there are other people to do the thinking and math for you, huh?
Because the multiplication table for 0.25 is too complicated...
Next you'll need an enterprise-grade library for adding numbers together. Because sums are hard, man. Who's got time to remember what 1+1 is?
So you need 1-unit padding? `p-1`. Double that? `p-2`. Still not enough? `p-3`. More? `p-4`. I don't have to think and do math in my head like 0.25rem, 0.5rem, 1rem,... and IMHO, it's a big enough win for me to keep using Tailwind CSS.
The features that were introduced in recent versions are just about supporting more CSS features to be more complete and not about feature-bloat.