I assume that "small event" means "short (in time) event". Fourier transforms are not good instruments for detecting short events. You would be better off looking for such events in the time domain, or with a wavelet transform, perhaps, that is set up specifically to detect the events you are interested in.

That being said, if for some reason the events have to be detected in the frequency domain then I would keep the transforms short. The reason for this is if the event is short with respect to the transform length it will look like a non-ideal impulse, which basically just creates a lot of high frequency content. Because of the small amount of energy (due to the short time) it will probably be difficult to pick the signal out of the noise.

If the transform is short, on the other hand, then it's energy is a significant part of the total transform energy. That will make it possible to pick the signal out of the noise, or even enable detection by measuring the total energy.

I assume that "small event" means "short (in time) event". Fourier transforms are not good instruments for detecting short events. You would be better off looking for such events in the time domain, or with a wavelet transform, perhaps, that is set up specifically to detect the events you are interested in.

That being said, if for some reason the events have to be detected in the frequency domain then I would keep the transforms short. The reason for this is if the event is short with respect to the transform length it will look like a non-ideal impulse, which basically just creates a lot of high frequency content. Because of the small amount of energy (due to the short time) it will probably be difficult to pick the signal out of the noise.

If the transform is short, on the other hand, then its energy is a significant part of the total transform energy. That will make it possible to pick the signal out of the noise, or even enable detection by measuring the total energy.