Due to the changes in Microsoft Jet 3.5 regarding the FlushTransactionTimeout registry setting, it is possible for throughput to decrease significantly due to concurrency issues when online-transaction processing (OLTP) is being implemented from many workstations. The reason for this is that Microsoft Jet 3.5 saves modified pages in its cache until no activity is encountered or the cache is full. Thus, the fast throughput of OLTP statements causes Microsoft Jet not to flush its cache and release the locks on index and data pages. This may cause many pages to be locked during this time period, which can cause concurrency issues when many users are trying to update one record at a time in the same table. To avoid this problem, use explicit transactions for these scenarios to force Microsoft Jet to flush its cache when the transaction is committed.