Network slices face challenges in supporting dynamic requests from user equipments (UEs) due to the resource constraints at the edge devices. This may result in potential anomalies due to resource unavailability or latency spikes. Existing schemes are hard to apply due to the lack of flexible steering of UE requests and multiplexed provisioning strategies. In this paper, we present a novel request provisioning model and propose a traffic steering framework to prevent anomalies and reduce the cost of serving normal UEs. Specifically, we categorize the UE requests into stateful and stateless types and utilize a multiroute resource provisioning strategy to address the UE anomalies. Additionally, the framework incorporates bandwidth-aware route selection and load balancing across multiple routes to improve the service bandwidth for UEs. Simulation results demonstrate that the proposed framework effectively reduces both the number of anomaly UEs and cost compared to existing baselines.