Интересно читать в англоветке, как три выдающихся русскоговорящих товарища (CfB, mthcl и kushti) десятками сообщений переписываются по-английски ... по самой важной теме, кстати (Transparent Forging).
Well, I believe that this hypothesis is false. In fact, I've analysed those games in Section 5 (see, in particular (16) and the calculations around Figure 5), and the conclusion was that there is only a small second-order effect in the case when no-one has a very big stake.
Maybe you are right. My assumptions include cooperation of malevolent forgers (via pooling their stakes and computing power together). This assumption looks reasonable to me because pooling allows to split expenses
without splitting reward in case of a successful attack. Security of Nxt is based on assumption that the worst scenario may happen, this is why BCNext bothered
* with 90% attack protection.
---------
* - This may be an incorrect word,
in Russian I meant "заморачиваться"Yes, it's reasonable to assume that cooperation of malevolent forgers
may happen, to the point that the hold a big stake together. But I don't think that it should be considered 'typical' situation. And if all stakes are relatively small, then even infinite processing power won't help much (in the "branching process" attack one already assumes that the bad guy has infinite processing power).
Тоже не знаю, как перевести "заморачиваться" 
