> If you need to set font-sizes with !important you also don't understand the cascade
If law requires you to make certain text a certain size, just do so.
Without any more background information, this explanation for !important doesn't make any sense.
Today it's obvious that overriding a font-size with a user stylesheet is a bit of an extraordinary measure. You're unlikely to have any legal issues based on somebody shrinking your text using one. This may not have been obvious in the meetings in which CSS was being designed: it's plausible that I might think user stylesheets would be all over the place and have unpredictable effects, including shrinking disclaimers or whatever. On the other hand, putting !important in a user stylesheet could've seemed like an extraordinary measure even then.
But again, I wasn't there. I'm just making semi-educated guesses.
If the originally purported intent had carried through to the specifications, it adds up in a sense. If !important overrides indicated an explicitly set formatting/sizing for compliance/legal reasons, implementers touching it without any historical context would know at a glance what it signified and that it shouldn't be touched lightly without explicit approval. And when the implementer updates the stylesheets to use 24px text on a page based on direction from the CEO/designer and the element with an !important override stays at 26px, the implementer can point to the !important override as the cause and confirm if the person requesting the change has confirmed legal/compliance approval to change that one as well. Even if the parties involved in the change (the designer, CEO, or implementer) aren't familiar with the compliance reason for the override's origins, the fact that it's there is a known flag that changes can lead to potential violations and costs/fines and need to be properly vetted for approval.
It would also make it easier to go in and update the stylesheets to comply with changes in compliance/legal requirements, as you can easily search through the styles for !important flags and find all the spots that need reviewed for potential change needs.
That said, its use in practice is completely different than this purported rationale for the flag's existence, and !important is sprinkled so cavalierly in existing codebases to crudely but effectively get a desired effect that the above situation is impossible at this point and the original rationale is moot.
Someone predicting "stylesheets all over the place" would have been exactly right, though - practically every single news page on the Internet has questionably-sourced ads, Facebook like buttons, Twitter embeds, and chat/discussion/engagement widgets, all of which have custom stylesheets. The only saving graces come from frequent use of iframes for isolated embeds (which will become less common now with cookie restrictions!), stylesheets well-engineered by well-funded corporations to have every rule prefixed by a custom and dedicated class name for compatibility, and preprocessors and CSS-in-JS that would make that latter approach viable.
Now imagine a world where you're a site operator, !important doesn't exist, and you cannot trust that a site component provider will not install a global CSS rule that will break your legal or branding requirements, with no real way for you to work around it if they do, without unmaintainable specificity boosters for your own components. You simply won't use that site component provider unless it's well-vetted. Which means that there's significantly less innovation on the Internet, because "nobody ever got fired for embedding IBM CSS" becomes an actual thing to think about. !important becomes the escape hatch that the web needs.
You are fundamentally unable to force the user to render the text you send them in a particular font size, !important or not. What if they use a text browser and small font size in the terminal?
The legal requirement explanation makes 0 sense.
Though I'll admit that once you have your "mandatory" size style as !important you'd only have to look at the other !important ones to be sure, so there's something to it.
Of course there's lots of perfectly reasonable uses of !important now: dealing with other peoples' uses of !important.