It's important that agents don't change their mind too often. You don't want a case where an agent flip-flops between chasing 2 different players because they're right on the boundary of which is closer.
Inertia is a score adjustment that can be added to future evaluations of an action (in a context) once you've decided to perform it once. This makes it less likely that in future updates the decision changes. The inertia value reduces over time so that eventually the action score is unmodified.
You can specify the inertia value for each Action Def in a Brain Config.
Another source of indecision is if running an action is dependent on a condition which then gets changed by running the action, meaning that subsequent evaluations suddenly rule it out, and therefore that action is liable to be interrupted.
To make this clearer, imagine that we use the "Creature Mood" as an input value for an action of "Alert". We want to execute the "Alert" action when the creature is "Passive", and that involves playing an animation, making a sound, that kind of thing.
However, the Alert action probably also changes the creature's mood! Say it changes it to "Alerted", but it takes a second or two to run its course (because of animations). If that brain updates again before the animation is finished, the action is going to score terribly because the creature is no longer "Passive", so it will get interrupted and not finish the alert animation.
Now, you could set this Alert action to be non-interruptible. However, this is a blunt instrument; it's nice to be able to allow interruptions always, just to prefer not to unless it's urgent.
You coudl also delay changing the mood until the end of the action. That would work but it's a bit unintuitive.
SUSS helps you address this by having a "Score Cooldown". This means that when a decision is made to run an action, the score (including inertia) is considered the minimum score for that action, which then reduces over the cooldown time. So if that action scores lower than the minimum over that time, it'll be bumped up so that it won't get interrupted, until it reduces enough to be outbid. It just makes it easier to avoid prevarication in your agents - you can set it to 0 if you don't want it, but generally it makes your agents behave more naturally.