This is a common trope I hear but I'm not sure it's true any longer.
One of the reasons Google login requires JavaScript to be enabled is to fix flaws in non-JS HTML that allow for hijacking passwords. It's why they took a much harder line on simple auth a few years back; too many instances of people being tricked into handing their Gmail account to scammers.
My mental model here is that the browser's password manager autofills a password that the user doesn't even know, and without scripting or dynamic resources, it's not possible to exfiltrate that data automatically. Without styling, form buttons would have a standard look so it's harder to trick the user into submitting something without knowing (c.f. the recently posted fake captcha that tricks users into revealing the visited state of links).
Edit: also, in a modern browser, passwords are autofilled by the browser which knows where input is going and can't be tricked. You could make an argument about other sensitive information, but that sounds like it's the same as plain ol' phishing.
Unfortunately, there's no language for querying the client "Do you support CSS" / "Do you support frames" without CSS so Google is forced to err on the side of caution and reject browser that don't support JS (because they can query that).
On the other side chromium devs need to develop quirks/hacks for specific sites in order to support their bloated javascript hacks (:
I find it bizzare.
https://en.wikipedia.org/wiki/Framekiller
It's one of those situations where "This was mis-designed with insufficient eye towards how right bastards could use it, but the cat is out of the bag. We can't change the spec because it'll break legitimate uses of this technology, so this is the best option we have." Similar to how images were naively scoped to be embeddable from any web domain and then people invented pixel bugs, but you can't change the security model around image loading without breaking vast swathes of the existing web.
That's never been sufficient to keep bad actors out because end-users don't grok the URL bar.
This attack is superior to a mirroring attack because it uses the target site's own UI resources, so it looks very legit (no need to pay someone to monitor the UI of the target site for changes and rework your exploit to attack them).