Resample following bandpass filtering?

[rmodrak]

Bringing back some old questions--

Should we automatically resample observed data and Green's function following bandpass filtering? Since misfit evaluation scales with number of time samples, the speed up could be significant.

What are the effects on accuracy, especially if previous resampling was applied (e.g. prefilter during instrument response removal)?

When data are sampled very coarsely, do specialized interpolation methods become necessary? (e.g. Instaseis uses coarse sampling, speakes about Lanczos interpoloation)