Volumetric lightmaps really help the huge overhead of shadows in VR, but the adaptive density of them seems to be generated entirely from geometry. So with a static light source, with interesting shadow volumes being cast from far overhead, all that detail gets lost except right near the floor.
Ideally the density would be based totally on how much variation there is at each point in space, weighted by places movables are more likely (X distance above the floor, similar to sparse volume lighting samples). With static direct lights this would place most of the density at the edges of shadow volumes near the floor.
This could maybe be an option, as it wouldn’t be as useful for stationary direct lighting with CSM.
In the screenshow I’m using static direct lighting. You can see that all the density is focused narrowly to the floor and walls, and not the edges of the light shaft volume up in the air, where the greatest variation in indirect (since the light is static) lighting occurs.
As a quick hack, you could generate shadow volume meshes for static lights using the old techniques and just treat them the same way the algorithm is currently treating scene geometry. It wouldn’t be perfect with the penumbra stuff dominates a long distances away from shadow casters in static lighting, but it would be a good start.