Enable Continuous Membership Reconciliation for quorum queues

Quorum queues can end up under-replicated when clients declare queues
while not all RabbitMQ nodes are available (e.g. during rolling restarts
or parallel pod startup after a data-wipe upgrade). RabbitMQ does not
retroactively grow membership, so the missing nodes stay missing.

Enable CMR in the RabbitMQ 4.x config for multi-node clusters so that
quorum queues are automatically grown to the target group size. CMR's
trigger_interval (10s default) fires when a node joins, covering rolling
restarts. The periodic interval (60min default) acts as a background
safety net. auto_remove is set to false so nodes temporarily down during
restarts are not shrunk.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: lmiccini

internal/rabbitmq/configmap.go

@@ -137,6 +137,20 @@ func buildOperatorDefaults(r *rabbitmqv1.RabbitMq, IPv6Enabled bool, configVersi
 		)
 	}
 
+	// Enable Continuous Membership Reconciliation (CMR) for multi-node clusters
+	// on RabbitMQ 4.x. CMR automatically grows quorum queues onto nodes that are
+	// missing from the membership, replacing the need for manual
+	// `rabbitmq-queues grow` calls after upgrades or rolling restarts.
+	// trigger_interval (10s) fires when a node joins, covering rolling restarts.
+	// interval (60min default) is a background safety net.
+	if IsVersion4OrLater(configVersion) && getReplicaCount(r) > 1 {
+		config = append(config,
+			"quorum_queue.continuous_membership_reconciliation.enabled            = true",
+			fmt.Sprintf("quorum_queue.continuous_membership_reconciliation.target_group_size = %d", getReplicaCount(r)),
+			"quorum_queue.continuous_membership_reconciliation.auto_remove        = false",
+		)
+	}
+
 	// Prometheus and management bind address
 	config = append(config, "prometheus.tcp.ip                          = ::")
 	config = append(config, "management.tcp.ip                          = ::")