20 dark frames is rather excessive. Personally, I'd question the effectiveness this. As is shown in a
flat field experiment, there comes a point where the quantity of flat field frames combined into a master makes little difference. The same rules would apply for darks, if fact I've tested this with a cooled astro cam (STL11k) and have seen little difference when blinking two masters (one of 12 subs and the other of 20 subs). I typically now use anywhere between 10-15 subs to produce a master (be it bias, flat or dark).
For the DSLR user, I'd recommend taking dark frames on the same night as your light frames. If you've modified your camera with active cooling, then calibration frames (such as darks and lights) becomes easier to maintain and are obviously reusable. Though I'd question the accuracy of these mods. They are different to the dedicated astro camera that can keep a chip cooled to .5 degrees of the camera's cooling set point (very accurate). If you're camera is capable of this, then you can easily reuse your calibration files for sometime. Though a chips characteristics change with time so it is wise to refresh your calibration library often. I typically refresh in two month cycles waiting for a cloudy or full moon night to record the darks.
It should also be noted that you don't need to take dark frames at the same duration of your lights. If your camera is actively cooled you scale your dark frames to match different light exposures. This is successfully achieved through the use of bias frames.
Either Sigma-Reject or Median combine is an effective method of creating the master. Just becareful of the sigma factor if you're using the this algorthm. Finally, if there is a calibration process that is going to stuff up your final image, it will be with your flat frames. If your flat frames don't accurately depict the telescope light path, you'll introduce gradients into your "calibrated" light frames.