@fallenartist @makc3d You could probably load the cube map into Blender and rerender it using a camera with a shallow depth of field. You could have either a equirectangular output, or automate a predictable cube map output by keying six animation frames with your camera pointing in each direction.
If you can get the gradual anisotropy and variable radii right, you could also just remap to a equirectangular projection and blur that. Or, if you have a high-end computer with infinite memory, you could also remap to mercator and not have to worry as much about the anisotropy I think.
My personal favourite trick for solving this problem is to print out the cube map textures with an inkjet printer, paste them on your apartment walls, and then run around with a spray bottle until you’ve appropriately blurred every part. (Then take apart an old scanner, stick the sensor bar onto a paint roller, and brush that all over your walls to re-capture the cube map images.)