Very good points there, although I think making trial accounts would scream "Scam" to me personally, also being that trial accounts can't trade, that wouldn't satisfy the logs. Scammed accounts would rectify this but would obviously heat up the situation a hell of a lot more.
My brain just popped out another one though!
The scam could be also be pulled off without having to actually scam the mark (don't flame me yet, let me explain

)
If you think, you have a group of 4 working together.. thats a good 80% chance of winning anyway and the mark has no chance of finding that out. If you are willing to pay the mark on a large win then you still have a good chance of a profit (although you don't have to).
Say you have a 1000G stake, you have that 80% chance of a win. Win once and kick from party and invite next one. Pull the actual scam on the times that you lose and then you have an average of 80% 'satisfied customers' (Satisfied to the point that they haven't been scammed).
If Blizzard was to then investigate, those 80% non-scammed peeps should hopefully help to prove that the mark was just a cry baby and a bad loser.
( And Kudos on the use of the word 'shrill', first I've seen on this forum

)
[edit] Just realised my maths is slightly flawed.. oops.. You DO need to scam for this to be profitable, otherwise you are just gambling.
5/5 games 1000G each stake = 4/5 to the group = 4000G total profit... 1/5 to the mark = 4000G loss... so that makes evens.
So obviously you scam on the 1/5 to make it from evens to 5000G profit (1250G each... not bad for probably 10-15 mins work)