EC2 to S3 file uploads. How can I tolerate ec2 failure during upload?

I have a service that lives in an autoscaling group. A client uploads a file to the autoscaling group - ec2 machines - some work is done and then the files are transmitted to S3.

My question is, what is a good way to recover from a failure of an EC2 machine that is in the middle of a transfer to S3.