Hi Bill,
Chris has summarised pretty well.
You will need flats 'every' time you change your optical train - and that includes whenever you change filters. Be careful when changing back to an optical setup for which you may feel you already have an adequate library of flats - can you really be absolutely sure that you have duplicated the optical setup 'exactly'? If not, and if your flats then prove to be inadequate, you may not be able to successfully calibrate the lights that you may have worked hard to obtain.
Taking good Flats is as important as taking Lights - and both are more important than their associated Darks, which are themselves more important than any Biases.
Darks (and Biases if you cannot maintain an adequate library of Darks) have a pretty long 'shelf-life'. Even if you don't have a temperature-controlled imager, then all you need is a library of darks that contains enough data to cover the exposure times you are using, and the temperatures that were encountered during the imaging session(s).
As for 'qunatities' of each type of image, then you can never have 'too much' data (only an 'unmanageable quantity' of exposures), My preference is to always acquire the same number of calibration frames (of each type) as I have Lights - and I always set out to acquire as many Lights as the weather conditions permit.
However, if you are going to do NB imaging with a modified DSLR, then you are going to be facing even lomger exposure times than normal - simply because you will actually be running those elusive photons through two sets of filters, not just one. Those longer exposure times will make the acquisition of the sheer amount of raw data that you need even more of a challenge (long Lights and Darks, and much longer Flats and FlatDarks - for every filter-change that you perform).
But, the rewards could be even sweeter, given the challenges that you will have had to overcome!
