Situation: 1 Author, 3 Publish instances using a shared file datastore (binaryless replication).
What is the best way to run Data Store Garbage Collection? Should it be run on all instances or really just Author since really, author will house all data?