Years later I understand but Marin on page 326 of his second volume of the English gives 19.Qb3 e5 20.Nd2 intending a white plan of c5, Nc4, Rb6 and Nd6.
Following Tony's line 19 Qb3 Nc7 20 Re5 Qd6 21 Re4 e5 can white play 22.c5 with Nd2 to follow ? Should queens come off, f4 becomes possible with b7 perhaps being vulnerable afterwards, d4 becoming weak and Nc4 / d6.

Tony, this looks familiar somehow. Is it in Tal's own book of his games, possibly?

Nunns Chess Openings also gives 15.Rb1!+= (page 24, note 9 - btw. under the topic "English Opening") without further comments. 

Following Tonys line to 18...Na6 I tried to find a way where white may sac the exchange on b7 and then keeps the connected pawns but failed (Mr. Byte showed me clearly). 

There is also a weird comp-line putting the queen on e1 which I dont believe (19.Qe1 Nxb4 20.Rxb4 Qa5 21.Bc3 (argh))  in and finally we can try to play c5 (instead of Qb3 for example), but this grants the d5-square to blacks knight and should be fine for him.

After all the rook looks (or is?) misplaced and is pushed back - the tide is rolling for black, bad news for white. I would certainly prefer black after Tonys 15...Nc5! which may even lead to a ?! for 15.Rb5 in the future. All magicans have their tricks - unfortunatly we cant ask Tal anymore

Perhaps 10...Re8! intending e5 at some point is just strong?

As part of my research for the DW book I was looking at the Réti line 1 c4 e6 2 Nf3 d5 3 g3 Nf6 4 Bg2 Be7 5 0–0 0–0 6 b3 c5 7 e3 Nc6 8 Bb2 d4 9 exd4 cxd4 10 Re1 Re8 11 a3 a5 12 d3 Bf8 13 Ne5 Nxe5 14 Rxe5 Nd7 15 Rb5 which is supposed to be good for White, but hit a major snag: 15...Nc5!:



which forces 16 b4 axb4 17 axb4 Rxa1 18 Bxa1 Na6 19 Qb3 Nc7 20 Re5 Qd6 21 Re4 e5 with a Black advantage, as played in some obscure Russian game. None of the Réti books I have even mention this, preferring to concentrate on 15...e5 ("the consistent follow-up" - Dunnington) which leads to two nice wins, by Tal and Miles.
Does anyone know what is wrong with this line for Black?