Re: Windows in 1983?
"And finally (yes, I am known for long rambling posts, so there's light at the end of the tunnel now!), substantial parts of the core OS were written in assembler by the very people involved in the design of the processor it is running on. Howzat for speed freak satisfaction?"
The umpire waves you away. It was an advantage in 1987, and the advantage mostly worked out for a few years because the architecture barely changed, but then, when the details really did start to change, assembly language programmers started to feel the pressure to keep everything working.
I remember substantial panic about the StrongARM's introduction, and the woefully late 26-to-32-bit switch (upon the complete discontinuation of any hardware capable of 26-bit operation) added yet more panic. By that time, the benefits of assembly language routines for stuff like blitting operations on bog-standard graphics hardware, when everybody else's graphics hardware had become a lot more sophisticated, had sort of evaporated.
And anyone having experienced early versions of Impression (and even later "stable" versions) can tell you that assembly language may buy you a performance benefit, but it leads to diminished stability and a lot more work to deliver additional features. The achievements of Acorn and pals were definitely notable, but the "on the metal" attitude helped quicken their demise.