Unsafe_aggressive_sstable_expiration - Procedure to enable unsafe_aggressive_sstable_expiration

I am planning to enable this tag to solve an issue related to some SSTables that are fully expired and are blocking some other expired SSTables from being expired. I have looked for information on which procedure we should follow, but I cannot seem to find information. For instance: should I apply these changes in all nodes in a cluster and then restart the whole cluster, or can this be done on a one node at a time fashion? I am wondering if there can be some inconsistency issues if one node at a time is changed and restarted, as the configuration would be different between different nodes in the cluster. For the record, we are working with a 4 node cluster, working with Time Window Compaction Strategy and which has a replication factor 3 and Consistency Level Local Quoruml.