Function deposit_scale 

pub fn deposit_scale(active_count: u32) -> f32

Per-particle weight of the additive deposit, so the total light laid into the accumulation each frame is invariant to the particle count (ADR-0065).

The draw blends One, One into a linear accumulation and everything downstream to the tonemap is linear, so without this the figure moves up by exactly the count ratio: attractor_particles is 50 000 at Floor and 150 000 at Rich, and Rich therefore rendered the same preset three stops hot. ADR-0045 and presets/README.md both promise a tier changes capacity and not behavior; for an accumulating additive scene that was false, because capacity is the picture.

So a tier now buys what a capacity tier should buy — the same figure sampled three times as densely at a third the weight each, i.e. less shot noise in the same picture rather than more light.

At Floor the factor is exactly 1.0 by construction, which is why no golden baseline moves and why that is assertable on the value rather than inferred from pixels. A future tier with a count below Floor would put this above 1.0 and amplify shot noise instead of reducing it — bounded and predictable, but worth knowing before a third tier is added.