How many times are S3 objects replicated? Does S3 increase the replication count for objects that are being accessed by multiple clients to distributed the load and get higher throughput?

I’m guessing AWS’ logic for 11 9s of S3 reliability goes like this: most S3 regions have at least 3 AZs. If they can assign 99.9% durability in each AZ and then distribute the object across 3 AZs, you get 11 9s of durability (probability of loss across all AZs is 1/1000^3).S3 objects are stored by a consistent hashing algorithm on the key name to select a server to store the object. The number of servers used to store objects in a bucket is based on the number of objects in the bucket, not how “hot” or busy a particular object is.If you need high throughput on serving from S3, you should use Cloudfront, which does care about throughput, in front of S3.

As far as I am aware replication is only available using cross region replication between two regions. How S3 archives the 11 9s is anybody's guess but they do not change the replication based on the frequency of access.