This limits the total number of concurrent calls allowed in the node, intended specifically to prevent data loss during extreme traffic conditions. Rather than letting the node accept calls until it's completely overwhelmed and starts dropping data indiscriminately across all connections, an admin sets this as a deliberate admission-control ceiling so that once reached, new calls are rejected cleanly while existing calls continue to receive adequate resources. For example, setting this to 1500 on a node whose hardware starts degrading data integrity above roughly 1600 concurrent calls gives a safety margin that trades rejecting new calls for protecting calls already in progress.