Have you tried terminating the EC2 instances abruptly to test the resilience of your solution?

aws ec2 stop-instances \ --instance-ids i-12345678 \ --skip-os-shutdown