"Subject to Review" is a great little 30 for 30 short on how this works in tennis. It makes the same point as you: it's surprisingly difficult to determine whether a fast-moving object fits within a particular boundary. Even in tennis, there's about a 4 mm margin of error with Hawkeye. That's often the difference between in and out. Soccer has also had instances where VAR has had to determine whether a ball was in or out or crossed the goal line (it happened in both the men's and women's World Cups, and to the US women's team). The reality is, it's just impossible to know for certain, and there are certain balls that can be thought of as BOTH in AND out.
In baseball, with the variable strike zone, this gets even more complicated. Plus, you're dealing with a much more three-dimensional space than a tennis in/out line. The modeling just has to be very sophisticated, and it's not going to be perfect. I'm good with the challenge system because teams then theoretically have the ability to strategically call on the technology. I really don't like the idea of a full electronic zone, though. Some pitches are both balls and strikes (because you can't really know the precise dimensions of the strike zone and whether the ball passed through them), and I think it's better to have a human making that determination most of the time. Fix egregious errors, and use the tech strategically. That's the best you can do.