Have you ever looked at the chart of the velocity of a rocket?
Rockets accelerate from 0 to 17,000 mph in ~5-6 minutes. the initial phase is exponential.
(it's not useful being pedantic here, I think I've captured what people think when they hear the word "skyrocket"; it's intentionally used to frame the discussion)
Solid rockets are pretty much binary in terms of thrust (they are either 100% or 0%), so they would accelerate fastest at low speeds, and accelerate slower as they encounter air resistance, and then begin to accelerate faster as they move into thinner air. But the rate of acceleration is, to my understanding, going to be highest in the first few seconds when air resistance is negligible, and about the same when they reach space where air resistance is 0. Exponential velocity increases mean that the rate of acceleration has to increase, which isn't something that is happening with rockets.
A "skyrocket" implies something different than a "spacerocket". Anything going 17,000 miles per hour is in LEO and no longer in the sky. Using this definition we should look at "fox 3" class missiles as they are launched from the sky at targets in the sky. Hypersonic missiles are actually air-breathing, and not rocket powered so they are excluded. Those missiles have an acceleration graph that is basically linear, but they start out already travelling several hundred miles per hour at a minimum, they then accelerate using a solid rocket motor (providing roughly equal thrust throughout its burn) up to a top speed of Mach 4 or so. After that they then use momentum to reach their target, but generally only lose speed after the engine cuts off.
So I posit that "skyrocketing" is starting from a fixed base, rapid linear increase, followed by a gradual decrease.
Solid rocket motors aren't binary, they just can't be dynamically throttled(early KSP was wrong).
In SRMs, usually a hole is left in the center so that the entire length burns towards inside-out rather than bottom-top so to avoid shifting center of gravity, and the shape of the hole is chosen so change in circumference length matches desired thrust profile. This usually means a star shape; the total lengths of edges in the star is longer than what with a simple circular hole, but the points erode faster and decreases reaction surface.
I think the more useful visualization is that a rocket flies at constant dv^2/dt ~ 1.4, with velocity following an exponential. The first derivative being constant satisfies the criteria for an exponential something.
Now imagine making a chart that adds a 100x multiplier to that rocket's data! Only that chart now can be called "skyrocketing" then, because it dwarfs the rocket's original data...
The fact that more "skyrocketing" things exist doesn't invalidate that the word "skyrocketing" could apply to what's presented here, and it does accurately frame the discussion around the sudden, alarming trend of self driving car incidents over time.
The various causes of the data jump/spike/skyrocket/whatever are fairly discussed in the article, and based on what's there it seems reasonable to conclude that there are much more complicated forces at play here other than, "self driving cars bad". The chart illustrates the sudden change, but explicitly tries to provide a number of very different possible explanations for the data.
(it's not useful being pedantic here, I think I've captured what people think when they hear the word "skyrocket"; it's intentionally used to frame the discussion)