RadiativeSources

The receiver is treated as facing each source — a parcel is a lump rather than a plane, so there is no one normal to take a cosine against. What that costs is that a flat plate edge-on to a fire receives as much as one facing it; what it buys is not asking a consumer for a normal per parcel, which is a number most of them do not have.

What a radiative exchange is bounded by, and the reason it has to be bounded at all: the radiative term is explicit, so a parcel whose surface shell has a small heat capacity — a coin, a nail, a leaf — can absorb more in one step than would take it past the source that is heating it, and then radiate the excess back as T⁴ and ring itself to pieces. It was found by putting a copper coin on a hearth in demo/dev/chemistry.ts, where it reached 4,000 K on the first step and 100 K a minute later.

Nothing can be heated past the thing heating it, so this is the equilibrium the step may reach and not exceed — which is what an implicit solve would give and costs one comparison per source.

§14's scan has to be coarse or sleeping saves nothing: this is a range comparison per source and no view factor, no temperature and no occlusion ray. A source's range already comes from its own power, so a bonfire wakes a parcel a candle would not.