Good problem to solve, but this particular solution is a fast path to hell for everyone involved.

You just can't scale text size independently of layout and interface. The size of the text is fundamentally related to the structural layout of the page. The number of columns, the size of images, the relative placement of buttons and UI elements -- it's all inextricably tied to the size of the text.

Good news is that we already have a solution for this: responsive design, aka page zoom. Every serious site already gracefully handles a wide range of viewport widths. When you zoom in, you are simply simulating a narrower viewport width. This type of constraint and flexibility is already well tested. Zooming in makes the text bigger. And, zooming in makes the layout adapt to a single column when that's all that will fit. It all works harmoniously together, because we test and accommodate for all viewport sizes, which is the same as all zoom levels.

The proposal at hand to scale text alone is bad for everyone. Developers now have a geometric set of permutations to test. What about an ultra-wide viewport with large text? What about a small viewport with large text? What about a wide viewport with small text? It's so much that it won't make business sense to invest in all of the testing, and all of the design and implementation work to accommodate all of the cases. And so, it will be bad for end users who will set their text size to their preference, and then find that actually usability and readability are now worse.

In the end the answer is simple: when users set their text size to be larger in the OS, browser vendors should increase the default zoom in browsers. This is already how it works on Windows, and it is definitely the best path to happiness for all.