One is going to be stable in Rust 1.52.0, the next release of Rust. This will happen before the next kernel release, so it barely counts. It's also a documentation tooling feature.

One is related to symbol mangling, and so in theory could be worked around I'd imagine, but is landing soon enough they don't think it's worth it.

The final three are related to each other; two of them look like they're being stable pretty soon, and the third, while it's less clear on timeline, is one pretty small thing.

So, the answer is basically that the exposure here is pretty small. How much that matters is a social question as much as a technical one. While "unstable" is a binary designation, within it is a spectrum of unstability. "This feature has no path to stabilization" to "this will be stable in the next Rust" and all things in between. Maybe allowing some unstable things that are closer to the latter is acceptable, maybe not.