Really this is almost a non-issue to people who aren't perfectionists
Outages are going to happen. If the only time this year your service crashes is when everyone understands that half the internet is offline because S3 is down, then you've got a ton of air cover on this one with your customers, who are probably themselves struggling with S3 being down.
Far from being a rationale to get off of Amazon, this is a good rationale to be using Amazon for their services. If you're a small shop you can go the Amazon route and only get hosed when Amazon borks up the whole internet -- or you can try to set out on your own and then when you inevitably screw it up because you've got vastly fewer resources than Amazon at your disposal you have nobody else to blame.
There's this third option which people think they have which is that its simple enough that they'll solve it all, but inevitably the complexity of your software and systems will bite you hard some day or night no matter how smart you think you are (and likely the fact that you think you should be able to reach perfect operational uptime indicates that you don't understand the uncertainties and you'll be much less likely to succeed). And something that kind of aspie perfectionist IT people don't understand is that the company that spends enough resources to try to get a flawless operational record will get beat by the company that spends enough to get by and diverts the freed up resources to other efforts to capture customers. So if your UI hasn't been updated in 10 years but your uptime is 100% and your backups are flawless you're probably sinking in the marketplace. And if you had the resource to do it all, you'll probably be a Fortune 500 company like Amazon -- the rest of us have to made business trade offs.