Techniques are described for reducing the impact of failure of a primary and a secondary routing control unit within a network device, i.e., a double-failure of the network device. For example, a network device, such as a router, initially establishes a routing communication session between a primary routing control unit of the router and a neighboring router. The initial routing communication session has a first restart time in the event of a session failure. The router reestablishes the routing communication session with a secondary routing control unit upon failure of the primary routing control unit. The reestablished routing communication session has a second restart time that is less than first restart time. Upon recovery of the failed routing control unit, the secondary routing control unit renegotiates the restart time associated with the session to an increased value with the neighboring routers to which the reduced restart time was initially advertised.