Elastic Beanstalk environment becomes unavailable once a instance requires to be terminated

Hi, I have an environment on Elastic Beanstalk with Application Load Balancer listen to HTTPS and HTTP as well, we also have a auto scaling group with 6 to 10max instances.

The application is facing an issue that we're struggling to find, a segfault (we're on it) once it happens in an instance it get terminated and start a new ok. However during the instance replacement, the state of the enviroment becomes "Severe" and the application get off, stops responding.

We're looking for a solution for high availability, if we have 4 instances running and one requires to be terminated for some reason, shouldn't the balancer keep the enviroment live and available? What should we look for to have it in place?