You're right about the energy math, and it's worth spelling out why the beacon still works the way it does.
A fusion generator is energy-limited. Its output is capped at 50 MW, and the engine computes the available energy as fluid_amount × temperature × heat_capacity. At the base 1,000,000°C that's 25 MJ per unit, so 2 plasma/s hits the 50 MW cap exactly — and at 10× temperature it only needs 0.2 plasma/s for the same power. Consumption scaling with 1/T is the entire point of that entity.
The plasma beacon isn't a generator, though. It's throughput-limited. Vanilla's own plasma description draws the distinction: "A quantity of plasma also represents that amount of coolant that travels with it." The beacon needs a certain quantity flowing through it to sustain its field, and hands the same quantity back as hot fluoroketone. That's why the ratio is exactly 1:1, and why consumption scales with quality — a stronger field needs more flow — rather than with temperature.
Here's the part that matters for base building. A fusion reactor outputs 4 plasma/s no matter how many neighbours it has; the neighbour bonus raises the temperature "without increasing the amount of coolant used", as the in-game description puts it. So a normal-quality beacon costs a quarter of one reactor at 1M°C, and a quarter of one reactor at 10M°C. The bonus never changes how many reactors you need per beacon — only how much electricity that same plasma could have produced instead. And if you've maxed out your neighbour bonus, electricity is not what you're short of.
I did look at the alternative properly. The clean implementation would be to make the beacon's hidden converter an actual fusion generator, which gives exactly the behaviour you're describing, in the engine, with no scripting at all. It fails on power priority: generators are demand-driven and sit at secondary-output, behind solar. With enough solar in the network the generator throttles to zero — your beacons would stop working because you built solar panels. Not acceptable.
The scripted route (swapping temperature-banded recipes at runtime) does work, but it would make an already cheap beacon up to 10× cheaper, and it would specifically undercut the higher plasma cost that quality beacons are meant to pay.
So it's deliberate rather than an oversight. Good question though — it sent me back through the numbers to make sure.
