Because, as Jon pointed out, the glows aren't necessarily linear or stable.
Although a simplification, think about how this works: A bias frame contains the noise generated by the camera simply taking an exposure and reading out the data. A dark frame contains that same noise plus any additional noise generated by the longer term activation of the camera for the duration of the dark frame.
If, for example, you calibrate a four minute light frame with a four minute dark frame then it would remove all of the inherent, not time-related, noise in the camera and also the noise that built up through four minutes of use. Done.
Where a Bias frame comes into play is if you want to calibrate a light frame but don't have a matching dark frame. Suppose that you have a two minute light frame but a four minute dark frame. What happens is that the software will subtract out the noise in the Bias frame from both the two minute light frame and the four minute dark frame. It will then scale (aka optimize) the bias-subtracted four minute dark frame by dividing the remaining values of the pixels by two in order to simulate what a two minute dark frame might oughta look like if you had one. It will then subtract that "fake" two minute dark frame from your two minute light frame. In other words, the combination of the bias frame and the scaled four minute dark frame should have about the same noise as a genuine two minute dark frame.
If you have a light frame that has a longer exposure than the available dark frame, the process is similar. Suppose that you had a six minute light frame and a four minute dark frame with a bias. The software will subtract the noise in the bias frame from both the six minute light frame and the four minute dark. It will then multiple the remaining pixel values in the four minute dark by 50% to create a "fake" six minute dark frame. It will then remove the "enhanced" noise found in this fake six minute dark frame from the six minute light frame.
This scaling process generally works pretty well so long as the noise in the image is, in fact, linear.
Aye. This is a simplification, but not too far off.
On my ASI-183 Mono Pro (which may behave differently than other cameras) I will use bias frames and a slightly longer mis-matched dark to calibrate my flats and it usually works fine. My actual light frames, however, are always calibrated with a matching dark and no bias. I don't have any experience with using the Pedestal, but the large-scale rejection doesn't seem to matter on my frames.