Please go through the IS and TN, clustering guide.
In the TN Processing rule, you don’t specify the IS instance. You only give the service name to execute.
By having a DB based Repository, setting the TN properties and enabling the clustering, you can achieve the true clustering using the 2 or more IS and 1 TN DB.
In that case if you change any Rule or Profile or anything in TN DB using 1 IS, all the IS in the cluster gets refreshed automotically with the latest values from TN DB.
Also, when a document satiafies a rule, the corresponding service may execute on any 1 IS in the cluster based on load (if cluster is based on least pending request) or alternatively (if the clustering is based on Round-Robin).