Anecdote, since the plural of "anecdote" is not "data". It is suggestive though...
0.47uF 16V X7R 0603, Samsung CL10B474KO8NNNC. Withstands over 120V without breakdown.Tested in a LLC resonant converter. Variable range quite large, and variability doesn't go away when hot. By "hot", we're talking on the order of 20 VAR.
Startup of the converter is interesting. It drifts severely at first, as the capacitors heat up. It reaches equilibrium, seemingly somewhere below Curie temperature. Possibly the core temperature is over, and varying amounts of chip are depolarizing?
Even when hot, the adjustable range is quite good. Hm, didn't seem like there was a big tail after adjustment, but there should because of the change in reactive power? May've been masked by the large time constant I used for biasing.
Tim