So I decided to give that 99.96% 10-6 DDB machine at the Rio a serious try.
It's a $1 3/5/10 play, so I elected the least variance model by doing 3-play, meaning it was $15 per hand.
I played $4665 worth of coin-in (311 hands of three each), and lost $1595. Ouch! I only got a 65.81% return on a 99.96% machine!
While 311 hands isn't that many, I felt I did spectacularly bad, and cursed my horrid luck.
Then I got home and realized this wasn't quite so uncommon.
Double double bonus is a very high variance game. Make it a 3-play, and it's even higher variance.
The variance comes from the reduction of 2-pair from 2x to 1x, and making up for it with higher-paying quads. This ends up increasing the importance of hitting quads to a level I don't enjoy.
I calculated that the approximate return when NOT hitting quads or better is 79%. Yuck.
You also need to hit your fair share of full houses -- basically one for every 92 hands you play. That accounts for 10.86% of your return.
If you go 92 hands without a full house or better, your return is 68-69%.
So in my 311 hands of three at a time, I needed 10 full houses, and I got 2. I got zero quads or better.
This put my expected return at about 71%, and indeed I got about 66% -- close.
Also, since the tier credits are reduced (1 per $50), I only earned 93 tiers!
Conclusion: This game has too much variance, and would require 1.25 million dollars worth of coin-in to optimally reach 7 Stars. No thanks.