We had to restart some nodes to make it work again.
We are monitoring closely the situation, even if the status says resolved.

We are investigating.

The authentication system came back online without any perturbation. However, it is not load balancing anymore for that component.

We have disabled all the nodes from the load balancer except one.
The selected node seems to work as intended, but still don’t know why the others break.

As investigation we are trying to remove node from load balancer to know which one is breaking.

It seems that there is a network communication problem between some nodes. The issue is that we can’t investigate.