Further explanation about why side-effects should be limited to useEffect

There are multiple recommendations about when side-effects are okay and when they are not okay; it is difficult to memorize them all, so it would be helpful if some further explanation is added explaining exactly why useEffect should be exclusively used for side effects e.g.
https://github.com/reactjs/reactjs.org/blob/master/content/docs/hooks-effect.md#example-using-classes-example-using-classes

For traditional life-cycle hooks, the reasoning is that some of them might be called multiple times before the render is committed; is the same applicable to hooks?

Thanks!

p.s. This comment seems relevant:
https://github.com/reactjs/reactjs.org/issues/302#issuecomment-345445888

从链接的 Hooks Effect 文档部分开始，将其中对类生命周期方法的讨论与当前关于副作用和 useEffect 的指导进行比较。更新文档，清楚解释为什么副作用应放在 useEffect 中，并说明同样的重复调用问题是否也适用于 hooks；当无需记忆各自独立的规则就能理解这项建议时，即视为完成。