The RW and WORM persistence stores currently require synchronization on disk read and write operations to avoid a Java bug with concurrent file extension. This code should be refactored to use a read/write lock where reads and writes obtain the read lock and file size changes obtain the write lock. This will provide a no contention path for the vast majority of all disk operations. Note that root block read/write operations can also use the read lock. There is nothing special about them in this sense.