The more I read about these FETs the more they seem to behave like valves/'toobs'. A symmetrical output stage driven anti-phase each side is a form of Class B. If there's a non-linear section at turn-off/on then bias above that towards the linear Class-A. That means a small(ish) 'resting' current at Class AB. I'd go for the 10mA/side 'static' bias. 100mA is operating each side in Class A - very linear but very inefficient (more heat!), fine for the Bitx40 but not the uBitx. Has anyone investigated the 'valve-type' transfer curve(s)? How much bias to get beyond the non-linearity into linear (distortion-free) operation for spectral purity?