I don’t have any experience with CMOS (yet) but with my CCD cameras I redo them every few months. Bias frames take no time at all but getting enough long exposure darks eg 1200 sec can take all night so I just leave the camera running on a cloudy night. I then produce a defect map from these and scale the master darks depending on exposure. I find the only significant changes have been with hot and warm pixels (and with one camera a faint line defect appeared) These are not a disaster as a cosmics removal tool normally zaps them but it is better if they can be fixed at source using an up to date defect map
With spectroscopic flats it is important to redo them if you move the wavelength range. I normally do them for each observation with the LHIRES and once a night with the ALPY. it takes little time with the built in calibration units though the results with the LHIRES at the blue end are very suspect due to light leaking past the slit.
I aim to sum at least 20 exposures (30 for ALPY flats where the light level is very low at the blue end)