Thank You! We used to get questions like this when I was coaching skydiving. "what happens if your main fails and your reserve fails". The answer these people are ultimately looking for is;
You die (or end up wishing you were dead).
Happy?
There have been a couple people who bounced and fully recovered. Of course they make lottery winners look unlucky...
There are two issues here with regard to diving.
1. Planning for double failures is pointless.
You will quickly drive yourself crazy and limit your penetration to about 20' if you start planning for multiple failures. It makes a lot more sense to focus on gear maintenence and prevention/avoidance of circumstances that may lead to multiple failures.
2. In many cases diving thirds is not conservative enough.
Consider what happens if you have a failure at the point where you reach thirds and are turning the dive that results in the loss of 1/2 your gas (a blown burst disc or extruded neck o-ring for example). Assuming you isolate immediately and save half the remaining gas, you have enough to get out using your own reserve gas if and only if you can get back out on the same amount of gas you got in with. If you are diving a siphon or if the cave gets silted out that could become problematic. If you were skip breathing on the way in to extend the time before you reached thirds (a really stupid thing to do in my opinon - get bigger tanks or get a stage) and are stressed going out, it could also be problematic. In any of those cases, you may end up using some of your buddy's reserve on the exit.
Deco adds a whole other dimension, particularly if it is unplanned and is a result of a silted cave and/or slower exit due to gas sharing with a buddy. You and your buddy's remaining gas needs to be enough to get you both out and through any required deco.
So depending on the conditions and a wide variety of factors there will be times when you should plan to turn prior to reaching thirds to hedge your bets.
Once you get into intentional deco, your gas planning and lost gas contingency planning gets more complicated and a conservative rock bottom approach makes a lot more sense than "x"ths.