It was mentioned somewhere in the DDoS thread that, instead of cycling coolant between the reactor and a cooling tower, spent coolant could be destroyed and new cells made from UU-matter. I decided to give this a shot.
As a test reactor, I used this design. 16 coolant cells, the same number of quad uranium cells, 1920 EU/t, 30 seconds before things start melting. Before all the coolant cells melt simultaneously, in fact.
Based on the facts that the reactor burns 16 coolant cells every 30 seconds, and that one coolant cell requires four tin (I'm using buckets of water to craft coolant cells, not water cells, which saves a little tin), and that the standard amplified Mass Fabricator cost of one UU-matter is 166667 EU, I calculate that the reactor must generate at least 5333 EU/t, not even counting the energy required to generate scrap (or other amplifiers) for the mass fab, or that required to smelt the tin dust into ingots. Not even close to what this reactor provides. However, I didn't actually go all the way through with the calculations (without errors) until I began writing this post.
A few miscalculations and similar sorts of confusions on my part thus led me to believe that this system might actually work, so I went ahead with setting it up in a test world.
As I am using GregTech, I initially used the Matter Fabricator with its 16666667 EU per UU-matter in place of the Mass Fab, but I quickly realized that it wasn't going to come close to working. So, I switched to the Mass Fab. Cheatery, I know, but it had to be done.
The setup included several RedPower relays feeding fresh coolant cells into the reactor as they melted. This leads to no heat buildup, if everything works properly- but even if some heat does escape into the reactor hull, the Overclocked Heat Vents take care of it rather quickly. And a Nuclear Control setup shuts off the reactor automatically if something does go wrong.
I found that, as the math would've predicted had I done it correctly, the buffer of coolant cells decreased over time, which would eventually lead to the reactor running out of coolant.
In short, it seems that coolant cells make for very poor single-use coolant. This strategy would only be worthwhile if you can manage to get more than four times as much EU out of a reactor as heat. No CASUC reactors here.
That said, if anyone can come up with a way to actually generate EU using these methods, I would very much like to see it. I have a feeling that it would consist of a modified recent-generation DDoS reactor.