If we take a frequency comb with finite width and spacing $2\omega_0$,
$f(\omega)=\sum_{n=nMin}^{nMax}\delta(\omega-\omega_0+2n\omega_0)$
the expectation is that this would be some sort of repeating pulse in the time domain. I suspect that the duration of each pulse would be something like
$\tau=\frac{1}{(nMax-nMin)2\omega_0}$
with a repetition rate of $T/2$ where $T=2\pi/\omega_0$. However, this is only by intuition. How would I go about formalizing the argument?
Try multiplying the series of Dirac impulses in frequency space with a certain window function. For example, this could be a rectangle between $(2 nMin - 1)\omega_0$ and $(2 nMax - 1)\omega_0$. However, e.g. a Gaussian shape is more realistic. Then, the signal in time domain is again a series of Dirac pulses convolved with the inverse transform of the window function, i.e., a chain of pulses whose shape is determined by the window function.
The Fourier transform rule for the series of Dirac impulses gives the relation between the spacing in frequency and the repetition rate. Furthermore, the relation between broad spectrum and narrow pulse shape becomes apparent (compare e.g. the sigmas of the Gaussian in time and frequency domain, respectively).