I ran into the same issue and was fairly surprised this was the case.
Our business analysts run incredibly inefficient queries via Atlas BI Connector. Isolating them to an analytics node should not impact autoscaling for our electable nodes. A single long running query does not mean we need more resources.
Our electable nodes are used for reads/writes by users of our application. This makes sense to autoscale because it will impact the user experience.
I ran into the same issue and was fairly surprised this was the case.
Our business analysts run incredibly inefficient queries via Atlas BI Connector. Isolating them to an analytics node should not impact autoscaling for our electable nodes. A single long running query does not mean we need more resources.
Our electable nodes are used for reads/writes by users of our application. This makes sense to autoscale because it will impact the user experience.