I remember something from my college engineering courses about thermal expansion.  I guess I should have paid more attention.

The fused quarts glass is excellent for this job because of its ability to handle very high and very uneven heating.  It does that because it has a very low coefficient of thermal expansion.  As it is heated, the hot part doesn't get bigger than the cooler parts next to it and cause it to break.  Stainless steel, on the other hand, has a coefficient of thermal expansion roughly 60 times higher.  Because we glued the tube into the flange with ceramic that doesn't flex, we ended up with the flange getting warm and breaking the glass tube.   

After much searching and shopping and more than a few visits to wikipedia, we learned that higher temperature (500C) , positive pressure applications (100 psi) of glass sealed to a metal flange is an extremely rare region to work in.  There are special types of glass where they gradually change the grade of glass from one type to a type that expands at a rate compatible with low expansion alloys.  The problem is, that special graded glass looses it's pressure holding capacity at the temperatures we want to work in.

The solution we settled on was to machine a deeper groove in the flange to allow for a thick silicone o-ring.  The o-ring allows flex, and being round, can also roll a bit as the metal expands and contracts. The trade-off is that the silicone is only rated for 260C, which will limit the ability to run this experiment at higher temperatures.  The flanges should run cooler than the glass and the center of the test cell, anyway, so it is good enough to let us replicate the conditions that Celani had in his demo.