Maybe my perceptions are out of date, but cleartype, while technically allowing higher font dpi, had a rainbow/screen door effect that stood out like a sore thumb for many of us.

The effect depends on the screen technology you use. Windows guesses the pixel layout, but you can tweak ClearType to fit your screen better. You can open a wizard that takes you through six steps and asks you to select the rectangle with the best text (out of six rectangles) to tweak it to your liking.

When the subpixels don't align, ClearType can be ugly as hell, but for me the default guesses always worked out.

On displays with high enough resolution/dpi, maybe. But isn't it just inevitable that if there are red subpixels on one side of your text, and blue on the other, that it would be noticeable?

I wonder if virtualization software messing with how ClearType assumes subpixel layouts could cause a lot of the dislike for ClearType in the Mac crowd, as many Mac users possibly predominately handled Windows in things like Parallels? I don't have any knowledge about how ClearType behaves in software like Parallels so I could be entirely off-base with that idea.

The only time I had dislike for ClearType was when I'd rotate a monitor from landscape to portrait. This makes sense, since the subpixel layout is then wrong. So, I'd usually just redo the ClearType settings when rotated and use the monitor in that orientation for a while instead of constantly switching.

The virtualisation problem may definitely play a role! In my experience, ClearType in a VM is often very very wrong. I think it may even be off by default if Windows can't find details about the current screen.

Alternatively, you may be getting the "RDP" experience for optimized drivers, which is just very blocky.

I prefer ClearType over whatever macOS does (or did, been a while since I had a Hackintosh) but combining different rendering styles is always jarring. Helping Mac people through remote desktop always made the text look wrong to me, probably for the same reason. When you expect text to look a certain way, alternative renderers are always wrong.

This is also something I notice with some custom GUI libraries that implement their own font rendering. I've dropped several GUI tools because I just got frustrated by how their text always looked different from normal text.

It can still be seen with the correct subpixel layout.

In my observations it’s the reverse. Windows font rendering looks ok to me (if a bit “off”) on my laptop with 1.5x UI scale, but underwhelming at 1.x. At 1.x fonts look badly kerned and a bit misshapen, and that cleans itself up quite a bit at 1.x and beyond.

macOS and Linux rendering looking better to me regardless of render scale. Good kerning and no beating font glyphs into odd shapes to fit a pixel grid on screens with pixel density high enough that it isn’t necessary (we’re not running 1024x768 LCDs any more).