The need for application-level context visibility to properly perform streaming service management in wired-wireless integrated networks is widely recognized. In particular, the paper claims the need for full application-level awareness of context data about the IEEE 802.11 performance anomaly, i.e., when even a single node located at the borders of the coverage area of a Wi-Fi access point produces a relevant degradation in the connectivity quality of all other nodes in the area. We propose a middleware that on the one hand portably predicts and detects anomaly situations via decentralized and lightweight client-side mechanisms and, on the other hand, exploits anomaly awareness to promptly react with application-level management operations (streaming quality downscaling and traffic shaping). In particular, the paper focuses on how our middleware performs anomaly-driven quality downscaling both to preserve the goodput at nodes in well-covered areas and to minimize quality degradations at the clients generating the anomaly. The reported experimental results point out how anomaly prediction/detection can relevantly improve the effectiveness of streaming downscaling, thus allowing to maintain acceptable service quality notwithstanding Wi-Fi anomaly occurrences.