If there's a problem with an aws service then aws will fix it. Of course there's always this weird bug that you will run into even in their services, but at the least, if there's an API that's documented, it works as documented for the documented scenario.

An autoscaling group in AWS might have issues if you complicate it a bit and run into edge case bugs, but otherwise it's most definitely going to scale your instances up and down (because they bill you to do so)