As I understand it, the trick is to learn how many textures you can combine in as few rendering passes as possible. Some hardware supports in a single pass just two color bitmaps being combined by a grayscale third, with that third often being vertex colors instead of a bitmap. Some will let the artist combine more textures at once in a single pass.
As I see it, it all comes down to how long it takes to render. If I have to re-render the surface multiple times (rendering passes), after a certain point the framerate starts to suffer.
Some engines don’t support 3ds max’s Multi/Sub-Object material type, a few do. If they do, each sub-material will be split into a separate set of verts by the exporter, creating in essence a separate surface. The verts along the seams between these chunks will get the same normals, so there is no visual seam in-game, but increasing the number of these chunks will slow down performance too.
The M/S-O material isn’t for blending bitmaps, it’s for dividing the mesh up into separate materials. If you want to blend, you use the map type they support, like a Composite map or a Mix map or whatever. It seems many engines these days set up their shaders in a separate app though.