You still have no control over obsidian itself. Any change can and will break plugins. So you either settle with one version for the next decade, or you have to maintain them. This is just the normal dependency-hell that every project has, where you have to compromise with external dependency and their whims. Just that neither plugins nor obsidian (to some degree) are the level of professional software-projects in that regard.

And let's not talk about changing Obsidian on fundamental levels. You have even less control on how it works on everything which is not accessible by plugins.

Depends on what you are doing. But the whole task & project-management-corner is constantly moving. Everything which modified the editor and preview was also regular breaking in the last years. For example, there were some plugins adding banners at the top of documents, or background-images or some icons. Or plugins modifying the yaml-area. They were all breaking multiple times when Obsidian was switching to the new live-preview-editor, then on changing frontmatter to properties, and on some other occasion IIRC. Usually after some months a new plugin appears, or someone forks the old one and fixes it. But as a user, it's pretty annoying to constantly have something breaking outside your control and getting stripped of features you want/need for various reasons.

Obsidian is useful, but far from being stable long-term yet. It's still very young.

> I could see the dataview plugin as an example

Yes, dataview was also very unstable the first 2 years or so, switching code and concepts, breaking old code along the line. It seems to be stable now, as the focus is on datacore.

> but most other plugins seem like they just add more convenient ways to do something that would be still pretty simple to do manually. (Templates for example).

Does it matter what a plugin is doing? If it breaks, it's a loss, whether it's crucial or just annoying.