Recently I had moved all my resources over to Amazon's S3. This morning, I woke up to a system wide outage that has beeing going on since at least 6:30am EST.
I was able to change all the asset code back to locally hosted by changing the "config.action_controller.asset_host" configuration setting and now all is working just fine.
Last year I lost a few Amazon EC2 instances and vowed to always have a backup plan that involved my colocated rack we could drive to. Although I have seen outages there, they have never lasted more than an hour. They also involved something that I could control and make reduntant going forward. With S3, there is not much you can except wait or have a backup plan.
Backup plans are fine for things like hosted assets but would be much harder where you have integrated customer data. Imagine if all your hosted attachments were in S3, there would be no way around an outage unless you mirrored the attachments in real time locally. That would involve paying twice for bandwidth and thus obviate all the savings with the S3 system.
I predict this will be big news and make a lot of people who relied n S3 think some serious thoughts about how they will structure their data going forward.
update 11am: There is talk the issue is resolved but it seems slow and it looks like there still might be some issues. To play it safe, I will keep the backup plan in place for another day to see what happens.