This 32 texture layer stuff sounds more like a semi-procedural texturing approach... Instead of making separate textures for gun #1, #2... character #2, wall #4 etc., they might be using many smaller dirt textures, scratch textures, metal textures, dirt alpha masks, etc. and mix them together to create various effects. Kinda like simple detail maps are used to boost up lowres ground textures, but on an even larger scale. For example you can have a detail in a grayscale texture which you use to multiply a low-res color texture, so that you can have the same thing in green, gray and brown versions. Then you add one or more dirt layer, using a tileable dirt texture but mask it with some fractal-ish or hand painted dirt map which you can also re-use for other models. Clever artists can hide repetitions and the possible variations are practically unlimited. Just take a look at the gun images, there's a lot of obvious re-use there.
I have to add that this is a very common practice in offline CGI texture painting, where you have to create very large res maps, 2k, 4k, even 8k for movie VFX models. The only way to do this is to keep adding various layers in Photoshop, all with some kind of 'dirt' masks; and use several levels of detail too, like 'small noise-ish detail', 'medium dirt', 'large stains of whatever'. The difference is that the texture artist has to blend all this together for the final image, which is then fed to Renderman for example - whereas in Killzone, they should be able to use the fill rate and computing power to do this 'flattening' in real time. It also has obvious opportunities for LOD, where you can progressively drop or replace successive layers of detail...
, but somehow I doubt that's what Guerilla was talking about when they boasted their "object detail" level management.
