Exactly. The only stable var and covar and determined to be so in hinsight. It's a function of performance. Either hire better traders, or structure the FoF equally with long volatility and short volatility strategies. There is no free lunch, eventually correlations go to one.
find the worst fund managers in each category and tell them to reduce their trading size to 1/10th their normal trade cap for your portfolio. then you suggest to them what trades to make. you should send me 1% of the profits each month for giving you the answer.
Essentially this is a mean-variance optimization problem. This problem is closely studied by the portfolio management crowd. However the key issue with this approach is that it rests on the assumption that the cross correlations between the strategies will be stable in the future. Another issue is that the exercise can only tell you the story of how the portfolio reacted to the specific macro economic forces acting during the calibration period. If there were no periods of high interest rates in the calibration period, for instance, then the results will tell you nothing about what will happen rates go back up again. Cross correlations tend to converge much more strongly than not for various reasons. When liquidity dries up in one market, it often dries up in other markets. Most strategies have a long beta bias, so true non-correlation is difficult to find. It might also be helpful to look at how each strategy has responded to a particular macro economic stimuli such as interest rate shocks, equity market shocks, currency shocks. This way, one can build up an intuitive feel for where the weak points for each strategy is and therefore for the portfolio as a whole. This lets you anticipate extreme conditions not present in the calibration time period.