I resolved after the recent flight-test failure to call it a day, tho' in view of the fact the equipment is in place and represents sunk cost, then I figured it's worth one more try.
Reverting to first principles I shall restore the outline of the frame to 36" from 30" and reconfigure the top-set of rotors (on the upper drone) the way they had been during a successful flight-test back in December of 2021: https://www.youtube.com/watch?v=HAIYGQ-CgQ8
To call on the rotors at the bottom end to participate in steering at this stage of the game is literally to stretch the elasticity of the airframe too far, though they could do wonders in terms of applying additional thrust. To this end, enlarging the airframe by those few inches provides for a tic-tac-toe planform that ~ in the hover at least ~ will be about as efficient as it gets.
You can ignore the overlap, incidentally, as the efflux from a propeller approximates to a funnel, so that by the time the airflow from the upper set gets anywhere near to the lower, it is unlikely to impinge at all. This effect I would occasionally observe whilst at the holding point of various runways, whenever the atmosphere was humid enough to show the vortices around wing-tips, propellers or the cowls of fan-jets.
It has then been, as the test-pilot said at conclusion of the last flight-test, a case of getting back to the drawing-board.
The pic below features the funnel-effect and that above the new configuration of each drone, with the upper in red providing cyclic control and the lower in blue collective.
