On the other hand, being stunned doesn't tend to be good for one's game even with plenty of time on the clock.

<24. ...Kxg7

25. Qg5+, Kf8

26. Qh6+, Kxe7

27. Qd6#>

... apart from control of the crucial d5 square and contributing to a healthy lead in development.

<... and practically let Black equalize.>

Having played 20...Qe8 perhaps Black should've tried 21...Bd8 (to at the least, remain consistent) and held his breath. He certainly had no time for 21...Rc7

Rybka supports the conclusion that the move is a very significant mistake by white, let me see if I can explain why.

First of all, it gives black the option of justifying his c7 knight simply by 10...Nxb5 = (Rybka) when white has not increased his development at all and will soon see the position truly is dead equal, eg. 11.Nxb5 a6 12.Nd6+ Bxd6 13.Rxd6 and I'm certainly not feeling nearly as confident with white as I was before his tenth move! How can he make progress against the stout black position with no weaknesses?

10...a6?! is a gamble by black. As indicated, 10...Nxb5 was the most obvious and probably best way to take advantage of white's slip. Now white can bail out with 11.Be2 when we must still nevertheless acknowledge 10.Bb5?! was a mistake because black now controls b5 which he can either use to expand his Queenside pawns to attack the white King, or place a piece on (eg Nb5) to control important squares in the white camp.

11.Ba4?! b5
12.Bb3 Na5
13.Qe2 Nxb3+
14.axb3 and look how black's position is improved! His Bb7 is free, white's Bb3 is gone, white's King's castle is compromised, and black owns most of the space on the Queenside around white's King!

You indicated that black's next three moves lead "nowhere near equality," thus justifying the entire 10.Bb5 plan, but this cannot be the case.

Look at the position after white's 14th move and black has done more than equalize, he is obviously better and Rybka confirms the advantage has switched hands (from .7 before white's 10th, to .3 before black's 14th). In material terms, white's questionable judgement has thrown away a whole pawn's worth of the advantage.

10.Bb5 was a very significant error by white, but black returned the favor when he played 14...Nd5?! instead of the most obvious move on the entire board 14...b4 when black remains very slightly on top after eg 15.Ne4 Nd5.

The exchange sac on d5 dovetails too well with K-side attack themes in Sicilian/Spanish (which means White already has several reasons to keep it high on his board-of-smiting). Deflecting e6 frees up canonical Ruy <21.Nf5>. Sicilian has the outright Nd5 sac just to get Nf5 (or, in some move permutations, you sac Nf5 to get Nd5, or offer two on d5/f5 and the one he doesn't get kills him). Theory was surely less prevalent back then (less complete, and much harder to disseminate, acquire, and study).

<Jesspatrick: ... stunned him with 22.Bh6> Well, only for 7-year-olds when their uncle first teaches them Ruy K-side attacks. Once White plants Nf5, deflecting g7 and Q-sees-g for mate is thematic in the position. That's <why> Ruy theory has Nf5 (and the laborious tours to get there), and Black defenses prevent it.