I think another 'flaw' with this is the fact that you will have 'cobblestone excess'... a.k.a, remains of rock from the supercooled lava that remain after injecting a ton of molten liquid into the system.
From what I can gather, it seems like you are harnessing the heat entropy of the cooling process from lava itself. I like this proposal, yet I find it quite lacking. How about this, instead of creating a meter-based system (with water levels and such), have a system similar to nuclear energy:
In this grid, you place in lava cells, much like you would with Nuclear energy.
W: Water Cell
L: Lava Cell
The cells, as they 'vent', emit and release heat at an alarming rate. At 10000 Kelvin, the entire chamber 'melts' (i.e. instantly becomes a lava block, and starts rapidly spreading).
After the cells have been 'used up', the socket is then replaced by about 4 cobble stone to represent the 'decayed' molten material.
Each water/coolant cell absorbs heat similar to the nuclear system. The only difference is that GeoGens would have a lower threshold for containing heat, as this heat is able to spread via conduction/convection. This is MUCH different than nuclear's radiation of heat (which transmits much differently), as well as the factors for lava cells multiplying their heat variable is much lower, as the material is constantly interacting with itself, and becoming one 'cell', rather than back-burning triggers like the atomic chain firings of atomic fission. Also, because the lava is much more 'free flowing' than nuclear, the cells are consumed much more rapidly than nuclear's process of energy management.
Thus the goal would be to get as much heat out of the system, while at the same time keeping the overall generator heat low, and thus EU generation maxed out.
EU Output: 10 Eu/t
Lasts: 2000 ticks
Heat: 250k per tick
Absorbs: 100k per tick
Lasts: 2000 ticks
Absorbs: 150k per tick
Lasts: 1500 ticks
Adjacent Lava cell EU/Heat multiplier: 1.5
(numbers can be adjusted for balance reasons)...