To accept the blackmail would mean to trust a partner who has proved to be dishonest. There is absolute no reason to justify a blackmail acceptance. Therefore that risk can be ignored.
Yeah, that's why blackmail is only a theoretical idea in the real world - right? 
Exactly. You should look into the whole Cryptolocker thing if you believe blackmail is to be ignored.
Ok lets play blackmail:
We do a trade and I (Bob Blackmailer) will not release the fund at the end of the trade, so you would loose 1.1 BTC and I only 0.1 BTC. I send you a message that I only will release it if you send me 0.5 BTC.
Would you accept?
If you accept to pay then you are in an even worse situation: You can lose now 1.6 BTC, I have already won 0.4 BTC. So I will try it again and will blackmail you. Would you be so irrational to pay again.... Then you would be in an even worse situation, and so on... after a few rounds I think everybody has learned that the only way to deal with blackmail is to absolutely reject it!
If you reject right from the beginning, what would happen? The blackmailer will wait a while, maybe repeat to blackmail you, but at some point he will give up. Then he has to choose to lose 0.1 BCT if he never releases the fund or he releases it and get back the 0.1 BTC. A greed blackmailer will release. And its in the nature of a blackmailer to be greedy ;-).
The only real blackmail problem would be in countries where BTC is illegal (russia) and the other party blackmail you with reporting you to the police. Also an undercover agent could be trading and sending you to jail... In those countries you have probably more problems then buying/selling BTC.
If you manage to do the Fiat transfer in an anonymous way (?) then that could be avoided as well. I have not idea how, but I assume in those countries they know ways to do that...
You are repeating your previous argument only with more words this time, and you clearly have not studied Cryptolocker. If you had done so you would know your assumptions are plain wrong, and that real life evidence of an already existing (and wildly successful) blackmail scheme contradicts every assumption you are making.
You see, the Cryptolocker guys are actually smart criminals. They did not, as you assume (and in doing so, greatly underestimate their cunning and intelligence), enter your supposed infinite ransom loop. Instead, they decided that keeping their word was in the best interest of their "business plan", if we can call it so. They established a weird kind of credibility, where the victims knew paying was the best option, bar none at the moment. And then they delivered.
Don't get me wrong, I very much like your idea, but you cannot go about dismissing the main objections by making generic statements like "...after a few rounds I think everybody has learned...", "And its in the nature of a blackmailer to be greedy ;-)." or "The only real blackmail problem would be...".
That is NOT how you do Exchange level security. These last few days should serve as a perfect example of why not.
There are tons of Cryptolocker case studies by very good net-sec firms. Please go read one to understand its severity and implications. Good will and logic assumptions do not make blackmail go away.