One practical addition for when you genuinely do need an effect and it is firing too often.
An effect depends on everything it reads while running, not on a declared list. So reading something inside a branch, or inside a helper function called from the effect, adds a dependency you did not think about. That is usually the cause of "why does this run when that changed".
The way to see it is to log at the top of the effect and then read the body carefully asking, for each reactive value: do I want a change here to re-run this? Usually one or two are accidental.
When you want to read something without depending on it, there is an explicit escape hatch for that, and it is worth using deliberately rather than restructuring the code around the dependency. The common legitimate case is reading current state to decide how to perform a side effect, when you do not want the effect to re-run just because that state moved.
But apply this after the previous answers, not instead of them. Reaching for the escape hatch to suppress re-runs on something that should have been derived is how you end up with an effect that works, cannot be reasoned about, and breaks the first time someone adds a dependency.