You might be wondering why you have to give S3 buckets globally unique names when S3 is a strictly regional construct. The reason for this is that each bucket is given a URL that can be accessed from anywhere, regardless of the region. There are a variety of strategies you can use to make sure your buckets are named consistently, and one of them is to use a variation of your domain name. For example, if your account is associated with www.example.com, you could prefix each of your buckets with com-example-www-. This would give you a reasonable chance of avoiding name conflicts.
S3 is a great example of the kind of resiliency you can get from Amazon's regional architecture that consists of multiple AZs, each of which a closely knit collection of data centers connected by high-speed data connections. Objects stored in S3 have 11 x 9s of durability, meaning they are 99.999999999% durable. The odds of permanently losing an object during our lifetimes are very slim.