We present a new RF fingerprinting technique for wireless emitters, which quickly adapts to the changing environment at the Edge because it is both real-time retrainable and power efficient. The quick and low-power training makes it amenable to the deployments at the tactical Edge, 5G Multi-Access Edge Computing, and portable spectrum sensors. While the existing work demonstrated high accuracy for hundreds of devices, flexible and robust Edge solutions do not exist. Efficient in-situ retrainability in the Wild is needed for RF fingerprinting to be used as an authentication mechanism in the Internet of Things. Our technique is based on augmented Ridge Regression (RR) classifiers, easily and efficiently retrainable on delay-loop outputs. The RR learns to identify emitters using bursts of emission samples, transformed and preprocessed by delay-loop reservoirs. Deep delay Loop Reservoir Computing (DLR) is processing architecture that supports general machine learning algorithms on resource-constrained devices by leveraging delay-loop reservoir computing (RC) and innovative loop trees, as demonstrated on our prototype platform. In prior work, we trained and evaluated DLR using high SNR device emissions in clean channels. We here demonstrate how to use DLR for IoT authentication by performing RF-based Specific Emitter Identification (SEI), even in the presence of fading channels and heavy in-band jamming by leveraging a matched filter (MF) extension, dubbed MF-DLR. We show that the MF processing improves the SEI performance of RR without the RC transformation (MF-RR), but the MF-DLR is more robust and applicable for addressing fingerprinting signatures beyond waveform transients (e.g. emission turn-on).