
If code that may remove items from m_nodes_by_id (Clustrix nodes keyed by id) succeeds, we must update the vector of health check URLs also in the case that code that _may_ add items to m_nodes_by_id fails.
If code that may remove items from m_nodes_by_id (Clustrix nodes keyed by id) succeeds, we must update the vector of health check URLs also in the case that code that _may_ add items to m_nodes_by_id fails.