> which runs on a completely dead CPU architecture that Linux doesn't even support anymore
The 68k is hardly dead; Linux support isn't a viability indicator of an architecture. It was dropped from Linux because there are no 68k systems powerful enough to run it and it's modern userspace, so it was a waste of development and debugging time. But the architecture is thriving in embedded and custom device spaces and fully supported by GCC and LLVM.
"Thriving" is a stronger word than I'd use. NXP still manufactures a limited number of ColdFire microcontrollers, which are loosely based on the 68k architecture. But they're listed on NXP's web site under "legacy MPUs/MCUs", and many of the parts are 10+ years old and NRND. It's pretty obvious that they don't plan on continuing the product line much further than required for support lifecycles.
They're legacy in the same way z80 and MIPS are. They reached an evolutionary dead end of full feature parity for their use space. The 68k hasn't been developed further because it does everything it's target audience needs and it's power user sphere was supplanted by ARM, x86/amd64 and PPC/Power.
Sometimes, a technology is good enough and doesn't need anymore. That's the entire reason the Cortex-M (and R) series exists; because you don't need a Cortex-A715 to drive a motor and monitor a thermostat.
Even then, there's still some sustaining engineering required to maintain a design and keep it relevant. The (e)Z80s that are being made today aren't the same as the ones that were being built in the 1980s -- they're being extended with new peripherals and ported to modern fabrication technology.
I'm fairly certain that hasn't been happening with the ColdFire series. Every ColdFire part I see listed on NXP's site is from 2010 or earlier, before the Freescale acquisition. This puts a lot of those parts 2/3 or more of their way through their 15-year availability commitment; if NXP intended to keep the line alive, I'd expect to see a lot more new parts, and that isn't evident here.
(MIPS is a weird one to mention because MIPS Technologies actually declared it dead last year and started trying to rebrand themselves as a RISC-V IP core provider. The main niche that architecture was used in was wireless routers, but that's been taken over pretty thoroughly by ARM these days.)
Don't get me wrong -- I cut my teeth on 68k and I loved the architecture. But it's also clear to me that it doesn't have a future.
The 68k is hardly dead; Linux support isn't a viability indicator of an architecture. It was dropped from Linux because there are no 68k systems powerful enough to run it and it's modern userspace, so it was a waste of development and debugging time. But the architecture is thriving in embedded and custom device spaces and fully supported by GCC and LLVM.