The one feature that breaks it for me for professional work is that filters happen to a layer instead of being attached to them an are able to be changed on a whim. It's impossible to do any real work if you have to render and re-render every shadow on every text and object when its size changes. I think I could do without everything else but that.

That’s why I love Krita so much – it can do all those things, can do animation, and, partially, already 3D, too.

For this reason I use Inkscape for most things I used GIMP for in the past.

The work you might have heard about GEGL - is all about providing foundations for such non-destructive editing. The highlighted split-view / curtain preview in the blog post is implemented using these new capabilities provided by GEGL. Adding new features that cement themselves as new things in the file format is part of GIMPs roadmap after GIMP-2.10 / 3.0. 2.10 / 3.0 are aiming to provide the same functionality that GIMP-2.6 / 2.8 provided but with GEGL as the core and thus higher bitdepth.

Photoshop has 'smart filters' and 'smart objects' which do exactly what you describe.

Is this mandatory nowadays? I mean it's nice that PS (and even Pixelmator and Affinity) have that, but I did design work many years ago and managed fine.

Sure it will be more inconvenient, but it's doable.

People did design work without any of these digital tools before.

Give people better tools and it's hard to go back. Can you imagine writing a large piece of software without autocomplete or syntax highlighting these days?

This is simply not true, Photoshop does have all those features!