While performing testing I needed to delete all of the studies in the archive (about 9,000 studies total). I set it so that studies that had not been accessed for one hour would get deleted. I also set it to delete the studies from the database. Memory usage slowly grew as the deletions occurred until it hit the 4GB limit for the heap. Some studies are larger than average, but it seems like this should not affect memory usage during deletions. The server was not handling any other transactions at the time.