Grabbing the signal off the PCB would require substantial reverse-engineering (or knowledge) of the machine. If it's a digital signal, you'd need to know all sorts of timing/framing information (it's not as simple as finding composite out, with all timings baked in). Chances are it's not even a single digital line but a parallel bus of some sort, so you'd need to find all the bits (plus the bit clock, and possibly frame clock) and know how to reconstruct the frame. Since these machines are using subsampling, you'd need to locate the busses for the Y and the UV planes and recombine them, etc. If the machine has a DAC, you could try tapping the signals going into the DAC, but that might have gotten subsampled to 4:2:0 (or worse), defeating the purpose.
On the bright side, a lot of these decks were used as part of larger editing setups. Often you'd feed data from two playback machines into an editing deck, which fed into a third machine (running in record mode) to store the edited output. For editing, I believe many of the links were digital, to avoid a reduction in quality at each editing pass.
I believe this was the case for D-3, but unsure about D-1. This was all way before my time.