I suspect that GPL2 was instrumental in guaranteeing that the work sacrificed into the common pot of Linux kernel is not going to be taken by a competitor when it's still unpolished, closed, and used to achieve market domination.

FreeBSD came before Linux (as 386BSD), and is also active used by the industry. How much code did Sony or Raytheon shared back to FreeBSD? (LLVM is not FreeBSD proper.)

See Android for how much that is working in practice, outside the kernel.

Or the Linux distros used by NVidia.

I will be honest mentioning Zephyr in a situation when talking about how outdated the Unix design philosophy is, is a bit funny to me since Zephyr (like ecos kinda did once) tries to be Posix-like in its APIs (but ends up not really improving things over the other embedded OSes TBH).

I am talking about Zephyr in the context of GPL, nothing else.

I find Zephyr to be a somewhat poor example. It's typically used on MMUless microcontrollers where the application is linked into the same binary as the OS. I'm sure you'll point out that it's not strictly necessary to use it in that manner, but that's how most people use it and that's how they expect it to work. Licensing it as GPL would mean that basically nobody would use it because it would require releasing your entire firmware source code, especially when there's other permissively licensed alternatives in that space like RTEMS, ThreadX, and FreeRTOS.

Exactly, there are no other FOSS kernels using GPL nowadays, the Linux kernel was the first and last one with commercial success.