K1LT's application is aimed at forming a decent beam to enhance received perfomance, that is the reason he uses 8 elements in 4 pairs of end-fire arrays. He has also added the ability to reverse direction to those elements. So yes, it would be frequency sensitive and have a relatively narrow range for a well shaped beam (and even at that you can see, in the QEX article, grating lobes affecting his beam shapes as he gets off boresite of the array). However, for RFDF and applying a null to QRM/QRN you do not need that complex an array (gain is not necessarily the goal, only a well formed null), and the system would be more forgiving of bandwidth. Don't get me wrong, you might still need several sets to get frequency coverage that might be most usefull, but each would be more simple than his setup.
For RFDF and QRM/QRN nulling a broadside array might be preferable.
T!