Hi there @FloraHyranoss !
I can certainly resolve some confusion (I hope. Let me know if what I write below is incomprehensible).
GI-1.0 is a ray-traced Global Illumination technique, that is to say, it's responsible for calculating bounced light in the scene. It's capable of handling dynamic objects because it's constantly updating those 25,000 probes using ray queries into the live scene. The snippet you linked to is specifically me talking about the software fallback we're developing in tandem, which uses some of the setup we had for single probes per room, and uses that data instead of hardware ray-tracing. We originally created that software mode so we could start prototyping GI before the ray-tracing engine changes were finished, but the results were promising enough that we decided to see if it could be used as a low-spec/no-RT mode, or possibly as a cheap second/third/etc light bounce.
Ray-traced DI is Direct Illumination, which instead of handling bounced light would replace/supplement the tiled lighting and shadow map system we use to apply the main scene lights. The leading solutions for this use a lot of similar concepts but the problem is a slightly different shape: with GI you're firing rays all over the scene trying to work out where the light is, whereas with DI you already know where the lights are, but you're trying to choose which lights (and where on those lights) you want to sample. Ray-traced DI can be useful for efficiently handling shadows on very large numbers of small lights, and can give much more accurate soft shadows compared to traditional shadow maps.
