So far I have received two kinds of flip-dot panels.

• A "PCB" type, where the coil is printed on the board. Very neat...
• A "wound" type, where the coil is made of enameled copper wound around the magnetic rod. Classic.

These two have different working points, as measured with a lab power supply:

• The PCB type starts to flip the dot when about 8V are applied to the coil. Add 0.7V (the unavoidable diode drop) and the LUMINATOR panel needs at leat 9V to operate. The nominal operating voltage is around 12V (compatible with my #YGREC16 - YG's 16bits Relay Electric Computer) but what's surprising is the operating current : the dot flips even down to maybe 20mA (the lab PSU goes in current-limit mode). I expected more... I suspect the PSU's output capacitors to store enough charge to create a current spike, so I'll have to test different configurations. Flipping power: less than 1/4W.
• The wound type (made in Germany) has a coil resistance of about 16.6 Ohms. The flipping current is around 130mA/2V, which means 3V or 3.3V with the diode. Flipping power: around 1/2W.

I'm amazed by the power efficiency of the LUMINATOR array. The printed coil is also more predictible and less subject to corrosion. OTOH the german displays are notably oxydized and several dots are damaged. I expected only one line (out of 19) to be unusable but mishandling might have occured and I can't get the 17 lines I wanted (16 data lines + parity, to spy on the memory).

Conclusion : if I give up on the extra parity line, the LUMINATOR is the best choice for my relay computer project. Thank you @Shaos for your precious help :-)

Now I have to examine the pulse characteristics...