Even for 10cm (~3000 MHz/3GHz) the high energy pulse that drives them has to be about a microsecond long. Tough to do when all your fastest electronic switches take about 10 uS to switch. So a slow switch dumped a charged up capacitor bank into (initially) a coiled up length of co-axial cable, shorted at one end. The pulse, reflecting off the abrupt impedance change (whatever 50 ohms to 0 ohms), went from long (10's uS) and short (600V) to brief (uS or less) and tall (many KV) and the average energy in the long slow pulse was transformed to the short fast one and dumped into the magnetron, peak power as it says, KW to MW. Lol then of course radar makes a huge outgoing pulse but then has to switch immediately into a sensitive receiver -- connected to the same antenna -- and switch fast enough to hear the weak pulse reflected back by the metallic target. So this sensitive hand-made diode is subjected to the KW/MW pulse... it's preptected by a clever TR switch tube that arcs internally, protecting the diode. It's all nuts, and developed by a small handful of very young british engineers. Pretty weird shit. Wartime or not its fkn magic. One engineer took their one working magnetron model to Bell Labs in NJ, flying on a passenger jety with the thing handcuffed to him and an armed gaurd, for Bell to figure out how to make them reliably/in quantity. MIT et al developed the shit out of them (NJ wasn't being bombed; London was) but it's entirely a british invention.