Tell me about a time you debugged a production Linux issue on an EC2 instance using tools like grep, sed, and awk, and how you managed the change across a fleet.
Write a function that parses a list of Linux log lines, extracts the affected instance IDs, and returns the fleet rollout order for a fix. Input is a list of strings, each line may contain an instance ID, a severity, and a timestamp. Output the unique instance IDs in the order they should be patched, with the most severe and earliest incidents first.
def rollout_order(log_lines):