Delay before considering the first of two redundant connections between applications or units as being down, and forcing a switch over to the redundant connection.