Hi thnks for your answer! Actually no, after log searching for a while, we did found out that the issue was the session handling.

To anyone with the same issue or similar issue, i will leave my experience below.

TL;DR: Filestore sessions where too many, and odoo GC got stuck, we set the flag  ODOO_DISABLE_SESSION_GC to avoid this

As our Database had for some reason a ton of sessions stuck, odoo sometimes make a garbage collections of sessions, and the way they do is, they ls the session dir (Odoo uses a filestore based session) and then loops over the entire files and rm each session file.

We used a EFS based filestore, so by coincidence, we tried to ls the session dir once from an ec2 instance, and that got REALLY stuck, and that was what happened. To solve it, we actually made our custom find --exec rm cron job that runs and erase all sessions older than a week, like odoo does, just that the way find does it is not blocking to odoo threads, and we relieved odoos responsability on that task. Now works like a charm we are handling a 50+ internal users and thousands of api requests :D 

Hey Federico,

It is posible that you forgot to add an inbound rule in the security group of your ECS cluster to allow incoming requests from your load balancer so that it can run the health check of your containers successfully. If your load balancer is not able to do a successful health check, it will reboot your containers. 

Look for the video o7s-eigrMAI on youtube at min 19:02 for the explanation. I wanted to paste the link but Odoo won't let me since I just created my account to reply you.

Best regards,
Daniel