AWS recovery of its highly available services like S3 from an AZ crash.

Answering your last question first - Do we lose any data? - S3 Standard has eleven 9s durability, which basically means if you have a billion objects you can expect to lose one object every hundred years https://aws.amazon.com/s3/storage-classes/#General_purpose

The answers to your other questions are linked - suppose your bucket is in us-east-1. This means every object in that bucket is replicated across a minimum of three AZs in that region (and as us-east-1 has six zones it's probably going to be replicated across them all).

When you access an object in a bucket normally, you don't know which AZ that copy of the object lives in, and you don't try and retrieve an object from us-east-1a or 1b or 1c and so on. All you know is you're accessing the object in a bucket that lives in us-east-1. So if an entire AZ goes down, the S3 service will just serve you out a copy of the object from another AZ that's still available. And you shouldn't face any increased time in accessing the object.