I've chased a lot of different system upgrades over the years to try to beat this. Even using NVMe SSDs to eliminate texture loading times hasn't really made a difference.
I'm led to the conclusion that the stutters have mostly to do with how the memory buffers were getting filled and then released. I'm sure that in 2000-2002 they did the best that they could with the systems available to optimize the process. With the advent of modern systems with plenty on excess memory, my attempts at monitoring it shows that this old game code never tries to take advantage of most the resources now available to it.
A few observations on my current system is that the micro stutters show up with their highest frequency when I'm going low and fast with the screen translating in the left-right direction (looking from the side at the outside of the aircraft). This may relate to their having used something like bit block memory transfers to shift every pixel a little across the screen. The hitch comes when they occasionally have to stop and redraw everything because the memory block overflows or gets too out of date.
When looking in the direction of travel, and the pixels are generally moving more slowly down the screen, the stutters mostly go away. Something similar happens as you climb in altitude. There are less detailed ground objects being managed, and the screen pixels are hardly changing as it's now just relatively empty sky. Also, it may not have to do directly with the display pipeline, and could just be the ground object memory management process, as reducing their density changes the frequency of micro-stutters as well.
Either way it feels like every time the cup overflows, and then gets emptied out, we see the slight pause in game rendering because even on really powerful modern computers the way they managed memory can't happen fast enough for us to not notice.
