Figure 9. Navigating from A to B on the navigation mesh
Many modern games do have high-quality pathfinding, and pathfinding often works well in some of the games shown here. But there are still too many games that do pathfinding the same way that games did in the 1990s.
To the best of my knowledge, most of these games use waypoint graphs for pathfinding. I think that's the reason for several of the pathfinding issues you see in this video, and many of the pathfinding problems we face in the industry as a whole. I believe waypoint graphs are now obsolete. This article explains the limitations of the waypoint graph approach and lays out a five-point argument for a better approach.
There was a time when it made sense to use waypoints for pathfinding. Back in the 1980s and 1990s, we were working under serious technical constraints and we had to cut a lot of corners. But we're a multi-billion-dollar industry now. Our target platforms have multiple cores and ever-increasing amounts of memory, and we can afford to do pathfinding correctly.
There's a saying among AI developers in the industry: "pathfinding is solved." We have good approaches to every pathfinding problem modern games face. We just don't always use them.
There's no reason not to have pathfinding that works in every game, every time.