I thought chemically inert and soluble were different. Either way the effect on the glue appears slow. I notice from wikipedia that Flourinert is highly soluble for oxygen for example so it could still be possible.
(if memory serves and memory is always unreliable)
The Prandlt number (
http://en.wikipedia.org/wiki/Prandtl_number)of FC-77 is half that of water therefore heat transfer performance of the fluid (as opposed to the system) is significantly lower as it is harder for the fluid to transfer heat to it self by convection, so you have to rely more on conductive heat transfer (which is worse but might be the dominant form in something like confined jet impingement; storm style blocks). Although since the response is non linear its effect on overall performance is small (think is proportionally to Prandlt^1/3 or 1/4)
1 degree in water cooling can thermodynamically speaking be pretty huge for changing one component.