I use a repeated cross-sectional data to estimate farmers' market participation decisions. I am including time and district fixed effects in my regression analysis. I currently cluster standard errors in districts (there are about 400), however what i do not understand is that under what circumstances one needs to bootstrap SEs?

I do understand that Bootstrapping allows assigning measures of accuracy (defined in terms of bias, variance, confidence intervals, prediction error or some other such measure) to sample estimates, but what i do not understand is that when does one needs to bootstrap SE's? and particularly in my case, since i am already clustering SE's in districts.

Thanks for your help!