Modern OSPF (Open Shortest Path First) routers can preserve their packet forwarding activity while they reboot. This enables maintenance operations in the control plane with minimum impact on the data plane, such as the Graceful Restart (GR) procedure. This of course assumes the stability of the network topology, since a rebooting router is unable to adapt its forwarding table and may cause routing loops. The Graceful Restart standard thus recommends to revert to a normal OSPF restart as soon as a topological change is advertised. This paper proposes to be less conservative and to take full advantage of the separation between the control and forwarding functions. This is achieved by new specific functionalities: (a) the prediction of routing loops caused by a restarting router, (b) the determination of the minimal number of temporary backup forwarding actions that should be applied to prevent these loops, without reverting back to a normal OSPF restart, and (c) the design of action plans to set and remove these temporary backups in order to avoid micro-loops when the restarting router goes back to a normal functioning. This results in minimal traffic perturbations when topology changes during a maintenance operation.