I have the impression there is little added value for more than 100 dB of attenuation in a stopband of significant width (i.e. not just a specific frequency). What is the point at which there is little practical application for more attenuation (80 dB, 100 dB, 120 dB, ...) over a stopband of significant width? Also what factor drives this point of no added value? My guess is that the highest useful attenuation is determined by the limited precision that we can measure voltage or any parameter that we might want to filter.
That really depends on the application. For example this processor: http://www.analog.com/en/processors-dsp/sharc/adsp-21364/products/product.html has 140 dB SNR in it's sample rate converter. That means the stop band attenuation of the interpolation filter must be at least as high but it's more likely to be 150dB - 160dB. You need extra head room to manage the fractional indices into the poly-phase filter bank in irrational sample rate converters