Networks operate under severe bandwidth constraints where environmental degradation, jamming and intermittent connectivity can reduce link capacity within seconds. Existing solutions to cope with such suddent bandwith degradation are suboptimal. For exam- ple, header compression is applied statically, congestion signals are consumed but not shared between nodes. No modular, extensible framework exists to dynamically compose and escalate compression modules in response to measured congestion while cooperatively synchronizing this state across middlebox peers. To address this gap, we designed and implemented an a multi-stage middleware that leverages Berkeley Packet Filter (eBPF) hooks at the XDP and TC layers to react to these congestions responsively.