To put some numbers around the problem, if one had a FT240 core which has
a mass of around 0.2kg, and specific heat capacity around 800J/kgK, the
energy to raise the core to a Curie point of 130°C to 300°C would be 17kJ
to 44kJ. If the core was well insulated (no heat loss) and dissipating
say 20W, that would take 15 to 40 minutes.

Of course, one would hope that the transformers do lose heat to the
environment, and that would substantially slow the rate of rise of
temperature.

Experience shows that a 5 minute test of a ferrite transformer does not
indicate continuous power handling capability.

Digressing slightly, but on this ferrite heat thing...

Martin questioned my article "A review of the Guanella 4:1 balun on a
shared magnetic circuit" at http://www.vk1od.net/balun/gsc/index.htm . In
particular, his issue was with my proposition that the extent to which
these things "work" is due to flux leakage on low µ cores, the lower the
µ, the more they resemble Guanella's balun.

Martin's inital experiments indicated that the thing did work, but on my
advice he tried prototypes on high and low µ cores and took thermal
pictures of the things after operation. The images showed non-uniform
distribution of heat in the cores which is either due to the main heat
source being the conductor losses, or that magnetic flux is a significant
contribution and not evenly distributed in the toroid. The flux
distribution is a credible explanation for the different patterns for
same winding on the different µ cores.

Again, this is one of those things that lots of hams have QSLs to prove
that they "work real good".

Owen