Thanks, Mike, for the great analysis in this thread.
1.  Rollout^{1}  21/11  eq: +0.2715 
 Player: Opponent:  68.72% (G:9.11% B:0.26%) 31.28% (G:11.54% B:0.26%)  Conf.: ¡Ó0.0024 (+0.2690...+0.2739)  [100.0%] Duration: 8 minutes 38 seconds 

2.  Rollout^{1}  10/6 7/1*  eq: +0.1807 (0.0908) 
 Player: Opponent:  64.76% (G:8.86% B:0.16%) 35.24% (G:9.29% B:0.15%)  Conf.: ¡Ó0.0030 (+0.1776...+0.1837)  [0.0%] Duration: 7 minutes 02 seconds 

3.  Rollout^{1}  21/15 7/3  eq: +0.1365 (0.1350) 
 Player: Opponent:  63.08% (G:7.14% B:0.19%) 36.92% (G:6.82% B:0.12%)  Conf.: ¡Ó0.0030 (+0.1335...+0.1395)  [0.0%] Duration: 7 minutes 11 seconds 

4.  Rollout^{1}  21/15 10/6  eq: +0.1139 (0.1576) 
 Player: Opponent:  62.87% (G:8.05% B:0.21%) 37.13% (G:11.18% B:0.20%)  Conf.: ¡Ó0.0027 (+0.1112...+0.1166)  [0.0%] Duration: 7 minutes 01 second 


^{1} 5184 Games rolled with Variance Reduction. Dice Seed: 7671034 Moves: 3ply, cube decisions: XG Roller

Above is Mike's rollout of the original position.
Below is a contrasting variant. (Sorry to use an XGR++ eval instead of a rollout, but I've been falling behind.) After playing A (Attack, 10/6 7/1*), the same four cracking numbers (44, 42 and 22) exist as in the original position  thought you might find it interesting.
Nack
1.  XG Roller++  10/6 7/1*  eq: +0.356 
 Player: Opponent:  71.65% (G:9.97% B:0.18%) 28.35% (G:7.08% B:0.05%)  

2.  XG Roller++  21/11  eq: +0.322 (0.034) 
 Player: Opponent:  71.01% (G:10.10% B:0.31%) 28.99% (G:10.78% B:0.12%)  

3.  XG Roller++  21/15 7/3  eq: +0.230 (0.126) 
 Player: Opponent:  66.41% (G:8.08% B:0.21%) 33.59% (G:5.70% B:0.06%)  

4.  XG Roller++  21/15 10/6  eq: +0.197 (0.159) 
 Player: Opponent:  66.43% (G:8.41% B:0.23%) 33.57% (G:9.71% B:0.08%)  
