You need a slicing system that understands breakover angle. That's not unreasonable. There are CAM systems which really understand the boundaries of the tool, cutting head, and machine.[1] The best ones can work in very constrained situations. It's expensive software, because it has to actually work right to avoid expensive damage.

There's a whole frontier here, to build 3D printing systems that have more intelligent path planning than just slicing.

[1] https://www.youtube.com/watch?v=st4MqJZobCM

I was thinking that (if possible) adding a sinusoidal vertical element to the layers (in both the x and y axis), similar to a crinkle crankle wall.

https://en.wikipedia.org/wiki/Crinkle_crankle_wall

There'll be a limit (some point before the head hits the previous layer), but it would still help to mitigate the weakness of planar printing.

I've felt that it's strange that these kinds of things aren't more standard, but I'm guessing it's harder to tune the machines for this?

The core idea is basically the same as that of the Nubian vault, so the core idea is very ancient. One can do this with snow too, to build somewhat surprising structures without having to use supports.

Back in the day I implemented this, the trick is that the firmware motion planner is generally by default not designed to handle smooth curves in 3D, only in 2D. We could have adjusted to fix that but the other issue was it makes the printing/slicing process even more unpredictable, which therefore makes it harder to use as an engineering tool.

The fact that the carousel doesn't stop on hover makes for a super frustrating reading/viewing experience.