Like Susan hints, if you use a setup with multiple, offset sensors, it requires very strictly timed read-out between the sensors.
In my mind the very legacy configuration for an RDF pops up as more practibly feasible:
2 loops or ferrite-bars perpendicularly oriented at 1 pole, with additional central pole-antenna possibly as global reference.
Simultaneous read-out of the signals by difference in amplitude should hint to angle toward lightning.
Minor detail that you have to find out whether lightning is in front or in the back.
Accuracy of such RDF-setup for radio transmissions is very reasonable, because stable mutual positions and transmission from point-source.
Accuracy for lightning-detection?
Probably not applicable, because lightning is electronically a very 'wild' event and also might spread over a wide area in the sky .....