Active connections still directed to disabled pool node

We have an https pool where we use a custom health monitor with the receive disable string looking for a 503. The ideal action is that if the monitor gets a 503 from one of the pool members, that pool member will no longer be used. However, we are noticing that users with previously established connections to that pool member continue to be directed to it due to persistence (despite the node being disabled).

We have the "action on service down" for the pool set to "reject" but any users with established connections continue to be directed to the disable pool member where they get 503 errors. Are we missing something? Is there some way to force the node offline instead of disabling it?