Whenever AWS is going to stop or replace a running container task the first send SIGTERM to the main process first. It is then the task of these processes to react and stop itself. This is part of the choreographed dance and if the process does not stop on its own it will be killed after a reasonable amount of time.
For our ECS deployments of Java Applications as a Backend this means that our Spring Boot Docker container hast to do whatever is necessary to stop itself. But what does this mean for currently running http calls or schedules?