KingG: Games like Polugaevsky vs Uhlmann, 1970 and Uhlmann vs Gligoric, 1971 more or less killed this variation. Radjabov avoids going straight into the type of endgame the occured in those games, but in the end felt he was obliged to due to White's control of the e-file(which is almost inevitable).
In the ...c5 Averbakh variation, it seems to be important to play and early ...h6 to prevent White from setting up the Qd2-Bg5 battery, which helps White to force play towards one of these horrible endgames. This is what Radjabov has done recently by playing 7...a6 8.a4 h6: Grischuk vs Radjabov, 2007 and Topalov vs Radjabov, 2008.
If White insists on 8.Qd2, then Black takes play into a type of Benko Gambit with 8...b5, Bareev vs Radjabov, 2007.
The immediate 7...h6 is also a mainline of course, even though it seems to have gone out of fashion recently for reasons I'm not entirely sure of. Perhaps it's because Black needs to sacrifice a pawn after 7...h6 8.Bf4 e6 9.dxe6 Bxe6 10.Bxd6, and although Black can usually hold the pawn-down endgames(Polugaevsky vs Kasparov, 1982), it's not exactly in the spirit of the King's Indian(and not much fun to have to memorize 30 moves of theory just to hold these endgames). White can also play the sharp 10.Qd2, which has given him favourable results, Li Shilong vs Y Vovk, 2006.
In light of this it's not so surprising that the more typical King's Inidan move 6...Na6 has been more fashionable in recent times, for example Grischuk vs Radjabov, 2007.