Gilmoy: It's almost a double-double! Run down the checklist:
- B thing <22.Be1> getting behind the Q
- R lift <24.Rxc3> getting in front of the Q; hence the offer <23.Nc3> is positionally strong
- Q step <25.Qd2> is not quite a canonical double-double, since the Q-R are not aligned (yet). But it still works because Rc5 will-expose an unanswerable triple on Na5.
The key insight must have been the brittle Qd8-Na5 lifeline. After <17..Qa8> Black is a marionette: <18.Ba6> with tempo, <19.Rc1 c6> strands the Na5 with <no escape>. Then <20.Bc5 21.Bb5 22.Be1!> make perfect sense: this N is a lasting target, and White can overwhelm it by force.
<10..Be6> is the novelty in this database, threatening (N,B)c4. <11.Nd2> meets it while maximizing the Bg2. <11..a4> seems to prevent the other defense b3, and is also a clearance for <12..Ra5> to maybe support Nd5. <13.Rb1> hidden-skewers weak b7, e.g. 13..Nd5 14.b4! anyways, and either the R retreats (which refutes the a4-Ra5 plan), or 14..Nxb4 15.Ba3.
Black's opening idea is deep and almost harmonious, but White unravels it with sharp play. Anyways, <13..Bf5(?)> is surely pruned from theory now: it does not solve Black's messed-up Q-side and the Bg2's pall. The only thing worse than moving it twice is to move it <three times>!