I am curious about the motivation for the fertility changes in 0.3.

In 0.2, neutron absorption between fuel rods resulted in a future increase in radiation (with a small increase in fuel cost) which meant that a reactor using lots of "fertility" would achieve a higher power, slightly more efficient reactor at the same size (barring the heat effects from this higher power). This was akin to the "chain reaction" that characterizes fissile nuclear materials.

The new system, as far as I can tell, means that using lots of fertility will decrease fuel consumption (with diminishing returns; I suspect that eventually putting neutrons towards fertility will actually lower the overall reactor efficiency). Furthermore, the fertility quantity is now tracked as a single reactor-wide value?

I'd like to know more about the thought processes behind this change.

It has to do with two things, usability and active cooling. Also, recall that BR operates on a "heat is bad" system, whereby the player wants to optimize their reactor for the lowest heat possible. In a real reactor the design goals are more complicated, but generally a real reactor wants to maintain a high target heat without going too high or too low.

With active cooling enabled, only heat is 'useful' energy. However, simply shunting radiation's RF generation into heat meant that active cooling would be, by default, considerably less efficient than passive cooling.

To simplify the logic, I therefore eliminated direct RF generation from radiation. Instead, all RF in a passively-cooled reactor is generated by dissipating casing heat, and radiation can directly contribute to casing heat. Active cooling simply replaces the dissipation mechanism with one based on a simple heat-transfer equation.

Second problem, usability. The original intent of radiation and fertility was to encourage players to experiment with different reactor designs. A player who understood how radiation and fertility worked could choose whether they wanted a bonus to energy output or to fuel-efficiency. For more power, let the radiation hit water or other blocks which provided energy when absorbing radiation. For fuel-efficiency, direct it towards fuel rods.

With the old system, fertility generated extra radiation. In the new system, this would have produced considerably more heat than before and led to weird situations like the casing being hotter than the fuel, which then leads to higher fuel heat, which eliminates the fuel-efficiency benefit.

The new system eliminates the negative-feedback loop while preserving the user-visible fuel-efficiency bonus, meets the original design goal, and is easier to understand. It's a sacrifice of some hidden complexity, yes, but complexity that a user can't see is generally more frustrating than interesting.

The tracking variables moved up to the reactor level so that a reactor's computational footprint while operating no longer grows linearly with the number of fuel rods.

Thanks for the explanation!