In order to prevent indefinite locking of resources, the transaction manager aborts in-flight transactions that have not completed after a specified interval, using a set of background processes coordinated by the TransactionReaper. The reaper rolls back transactions without interrupting any threads that may be operating within their scope. This prevents instability that results from interrupting threads executing arbitrary code. Furthermore, it allows for timely abort of transactions where the business logic thread may be executing non-interruptible operations such as network I/O operations. This approach may, cause unexpected behavior in code that is not designed to handle multithreaded transactions. Warning or error messages may be printed from transaction-aware components as a result of the unexpected change in transaction status. The transaction outcome should usually be unaffected. Any problems can be minimized by tuning the transaction timeout values. See Chapter 17, Datasource Configuration for more information.