Application of Ethernet-based networks in the autonomous and automated vehicles requires fail-operational behavior. The network components must not only detect failures, but also reduce their effects on the system’s safety, as it may not be possible to bring the driver back into the control loop. In this work we present a mechanism which, through protocol-based synchronization and fine-grained re-configuration of network components, allows achieving traffic isolation, fault recovery and controlled degradation of network performance. This enables the maintenance of (degraded) network operation in case of a malfunction, a safe stopping of a vehicle or its safe operation until the driver takes control. We offer a detailed evaluation of our approach in a demonstrator setup as well as a simulation and comparison with state-of-the-art approaches.