So I did some math on copper efficiency and gregtech lava centrifuging. Assuming you're using TE to manufacture lava for sustainability, and efficient electric engines (2.5 EU/MJ) to power the crucible, the effective output of a reactor should be reduced by .2525 for each copper in a thorium cell, .63125/copper in plutonium, and 1.2625 EUt per copper in a uranium cell. Higher consumptions may be necessary for non iridium reflectors (the tin output of the centrifuge will be evenly matched, coal is a problem).
Some results: the quad cell+4 reflector setup takes 212.1 EU/t to sustain, more than it produces.
The normally highest overall efficiency reactor takes 121.2 EU/t to sustain. Its gregtech overall efficiency is 2.02. So worse than 0 running cost setups.
The highest copper consumption hybrid (the 420) takes 7.575 EU/t to sustain, giving it a 1.8% automation loss. The thorium neutral takes 6.3125, for a 1.7% loss. Even when thorium life is halved the stuff should still be viable in multicell form.
Nether lava centrifuging should work out much better, though its not really possible to 100% automate that. There's also a question of lava efficiency, IE, is that lava worth more in a thermal generator. The 4 reflector setup would consume 2,688 buckets of lava every 10,000 seconds, equivalent to 80 million EU. That same reactor will produce just 28 million EU in the same period.
As for a hybrid, the thorium neutral reactor will produce 367 million EU, compared to just 3 million EU worth of lava.
Also: Peppe, I need to thank you yet again, the railcraft based reactor refueling system is incredible for multiple reactors, I've got it maintaining 5 reactors in creative right now. As best I can tell the system should be trivially expandable to any number of reactors.