In the process of cleaning up datasets we no longer use and looking at disk usage for an Enterprise geodatabase, I discovered that nearly half of the disk usage is coming from the historical tables of two frequently updated items, a feature class in a feature dataset and a table. These two items participate in a relationship class with each other. Between the two of them, the tables are taking up almost 95 GB, whereas the current data only takes up about 3 GB combined. Is there a way to keep the size of the historical tables in check without turning off geodatabase history?
There is whole help topic section on "history" that may be useful, check the table of contents to the right of this link
maybe something might be useful