Early attempts to make very high frequency oscillators used the Barkhausen-Kurz (B-K) mode. Any cylindrical triode will tend to oscillate in B-K mode, typically at 100s or even 1000s of MHz, if operated with the anode negative (eg. -200V) and the grid positive (eg. +400V) but this tends to overheat the valve and spoil the vacuum unless the valve has been previously processed to a very high temperature.
One way to process the electrodes to orange or white heat without destroying the glass and the seals is to make every electrode from a single strand of metal brought out at both ends and to pass large ohmic heating currents through each electrode in turn whilst the valve is being pumped.
Until the arrival of the magnetron, invented around 1922 but of little practical use until redeveloped during the early 1940s into the cavity magnetron, B-K oscillators were the only known means of generating CW at centimetric wavelengths. They were notoriously unstable and short-lived but they were used for numerous experimental purposes including early attempts at radar.
One of their technical limitations lay in low power conversion efficiency, in theory 8.5% max. but usually much less, and in the difficulty of coupling this power out of the valve via thin leads passing through glass seals. There was no technology in the 1920s to permit the provision of proper microwave ports, either coaxial or in waveguide.
Valves intended for oscillation in B-K mode have been obsolete since the introduction of the reflex klystron (c. 1940) but the majority of large cylindrical transmitting valves (including tetrodes) are only too ready to emit unwanted bursts of B-K energy if their anodes are allowed to run negative, even momentarily, while positive RF drive is applied at the grid.