[chipsa]

As sensors get better, the closer the penetration aid ends up having to have the same characteristics as a warhead filled RV. At which point you're better off just launching a real warhead than a decoy. Ex: Low orbit isn't free of air, just very very low density, so there's still some drag. Your decoy has to decelerate the same rate as the warheads (which is fractions of a mm/s^2, but still). Or the temperature of your decoy changes alot faster than the real thing, because it's light. So, they cool down faster, unless you add something to heat them up. But then we can also try to heat them up, and they'll still heat up faster than the real thing (space based laser or whatever to do this). And the newer sensors that we already have are imaging sensors, so they it's not just that the have to have the same radar signature, but it actually has to look the same to a certain extent (X band is ~2.5-4cm wavelength, so the features bigger than that need to be the same).

But yes, best to do it before the missile debusses the warheads. Which is partly the point of GMD: yes, it's name has mid-course, but part of the point is that it pushes the point between post-boost and mid-course much closer to launch, which results in the warheads being released earlier and farther from the targets, which makes them less accurate.

Another part of BMD that isn't commonly appreciated is that Aegis/SM-3 has sufficient performance to cover about half of the US.