@Joey thank you. What else can we investigate? We're running out of ideas. We observed a lot of write operations as @Pablo mentioned and the connections number went high and hit our limit. We observed it both in production and later when trying to reproduce on our dev environment. It doesn't seem to happen immediately after enabling this flag but some time later but we're not sure. As Pablo mentioned we only enabled it on our read instance. We'll appreciate any suggestions, we're kind of stuck here.