If the glitch is at the end of digital to analogue conversions, a clock pulse at twice the sampling frequency can be used to sample the output from the resistance ladder in the middle of every sample. This avoids the glitch altogether. For this, a sample and hold circuit could be used.
This is something, I never tried, but it is logical.
This is something, I never tried, but it is logical.