> can you guarantee the generated CSS doesn't put the browser's layout engine in an infinite loop or something?

the way TW aproximatively works is to collect a set of used class names from your code (e.g. flex, min-h-screen, bg-[url(/img/grid.svg)], absolute, etc.) and produce the CSS for those classes.

For the random 5 classes above it would produce:

    .absolute{
      position: absolute
    }

    .flex{
      display: flex
    }

    .min-h-screen{
      min-height: 100vh
    }

    .bg-\[url\(\/img\/grid\.svg\)\]{
      background-image: url(/img/grid.svg)
    }

By default TW also includes a CSS reset preamble and defines internal variables.

Sometimes the generated css contains complex selectors but as far as I know it is not a common occurrence

    .space-y-4 > :not([hidden]) ~ :not([hidden]){
      --tw-space-y-reverse: 0;
      margin-top: calc(1rem * calc(1 - var(--tw-space-y-reverse)));
      margin-bottom: calc(1rem * var(--tw-space-y-reverse))
    }

In other cases more css variables (custom properties defined in the preamble) are used

    .-translate-y-1\/2{
      --tw-translate-y: -50%;
      transform: translate(var(--tw-translate-x), var(--tw-translate-y)) rotate(var(--tw-rotate)) skewX(var(--tw-skew-x)) skewY(var(--tw-skew-y)) scaleX(var(--tw-scale-x)) scaleY(var(--tw-scale-y))
    }
    .shadow-xl{
      --tw-shadow: 0 20px 25px -5px rgb(0 0 0 / 0.1), 0 8px 10px -6px rgb(0 0 0 / 0.1);
      --tw-shadow-colored: 0 20px 25px -5px var(--tw-shadow-color), 0 8px 10px -6px var(--tw-shadow-color);
      box-shadow: var(--tw-ring-offset-shadow, 0 0 #0000), var(--tw-ring-shadow, 0 0 #0000), var(--tw-shadow)
    }

I believe that CSS performance is mostly dominated by size in bytes of the stylesheet, number of properties, complexity of selectors, and frequency of updates. I suspect that TW improves most of these metrics in exchange of bigger html class attributes.

https://play.tailwindcss.com/