There is a possibility to let nodes temporarily desync from the cluster when their queue of pending write sets is getting to large due to a large query executed on them. Normally a node will use a flow control mechanism to slow down the cluster when their queue of not yet applied write sets reaches a limit, but by setting the parameter wsrep_desync to ON, a node will instead disconnect from the cluster when under heavy load and try to reconnect and synchronize on a later moment with an IST or SST state transfer. The rest of the cluster members will continue to synchronize with each other without delay.
Please note though that if you use this setting, the state of your slave select node will lag behind the state of the cluster and this is not acceptable in all applications.