But this doesn't apply to the Roulette paper, as a constant flow of condensation would stop the cell walls from drying out
The calorimeter used in the Roulette paper is slightly different from standard isoperibolic calorimeters (IPBCs). First off, the usual assumption about IPBC's is that they temperature measured is uniform, i.e. no hot spots, etc. The Roulette calorimeter obviously is not uniform in that sense. The three temperature measure points show different temperatures at the same time, which is why the set of TCs used to calculate output power is dependent on input power level. Which set to use at a given point is an interesting question. In any case, I would have to study this more fully to decide how a CCS might occur in this cell (and I don't see the point since I can recall no other repot where this kind of calorimeter is used). As I recall it is an open cell so electrolysis gases are normally exiting the cell, and thus if they find a way to recombine (anywhere in the cell in this case) they could induce a CCS. Also note that since the set used depends on input power, the upper set may not necessarily be covered in condensate as you suggest. Their cell is reminiscent of a distillation column, which initially starts out hot at the bottom and cold at the top during the early boiling phase, but that then proceeds to heat up as more power is applied to the boiler.
But in the end, the one-shot nature of this report is what is important. There are many questions about how the cell/caloriometer functions, and no data to answer them.