I don't know if it's still the case, but not too many years back NASA was hoarding 486 CPUs for space missions, because they are an absolutely known quantity: ALL the bugs are known, so they won't surprise you with some new glitch a few million miles out, not exactly convenient for repairs or updates.
Until everyone who is familiar with those known bugs retires, or dies, or forgets, and you need to go find Clint Eastwood to fix your damn Soviet nuke satellite.
It's always a balancing act between what's well-tested and what's sustainable, between what works well now and what may work better in the future.
IMO, it's a horrible idea to rely on any toolchain that new developers aren't learning right now.
ETA: I don't care how stable it's been. If it breaks tomorrow, how easy is it to fix it? How easy is it to find someone willing to support it? If the answer isn't "very easy," then it's time to switch platforms.
Last edited: