I'm new at this. Why is C++ an obstacle to bootstrapping?

It's the circular dependencies.

GCC 4.8 or higher needs GCC 4.7 or higher to build - so do you come up with a way of bootstrapping GCC from itself without an existing C++ compiler (for example, Perl is self-hosting, but the bootstrap is done with 'miniperl' that can be built with C alone), or do you start with GCC 4.7 that you can bootstrap and go from there?

The glibc problem is more interesting because it's an indirect dependency. glibc needs python to build itself, but python needs glibc.

The point the article makes is that solving these problems is non-obvious and solutions come with tradeoffs.

...glibc needs python to build itself, but python needs glibc.

Surely there must be a glibc which predates the Python dependency. Why can they not build the prior glibc version and use successive releases to iteratively build up the platform?

Not technically satisfying, plus probably some turtles-all-the-way-down situation where there was a pre-Python dependency which is equally hairy, but conceptually, I do not understand the limitation.

Too late for the mountains of existing software, but I know Zig took some effort to address the bootstrap problem by relying upon a WASM seed.