In "short time" you can consider that the energy left
in the resistor film (which is not the whole body) is all
of the pulse energy and none escapes by conduction
(the adiabatic approximation). In this case you have a
power-to-blow which is greater than steady state, but
does not come close to duty cycle averaged (this
assumes conduction balances input power, which at
short pulses is not true).
You can do the calculations if you can get manufacturers'
information on the volume of the resistive material and
the temperature at which Bad Things Might Happen.
It could be that a thin (or thick) film resistor simply
evaporates and entrains to the the encapsulating layers
(as integrated thin film NiCr resistors do), it might be
reaching the material liquidus such that the "fuse"
just melts and pulls open, it could be some metallurgy
thing that simply results in an unacceptable (to some
criterion) drift in value or stability.
Joule energy of the pulse, resistor active volume,
material density and specific heat are all you need
to figure the temperature rise. The "fail" temperature,
the criteria and so on reside (if anywhere) in the bowels
of the manufacturer's reliability organization. Probably
all sandbagged to hell, in the spec you got.
But it's a real thing and needs to be respected /
understood / quantified.