key reasons to cool
1) management of dark shot noise
2) management of cosmetic defects/Dark FPN
3) management of RBI
There's a relationship between the level of cooling used and the maximum time you can expose and still have the dark shot noise less than the read noise of the camera. That time point is usually regarded as the maximum practical exposure time (see pages 6-9 of the linked pdf below)
http://www.narrowbandimaging.com/inc...ling_crisp.pdf
Cosmetic defects can and do accumulate with time; usually they are pixels that have higher dark signal than surrounding pixels. The bad effects of such leaky pixels can be minimized by deep cooling (page 10)
RBI is a problem to varying degree with KAF series sensors. The KAF09000 is particularly bad. The best known method for managing such RBI is to flood the sensor with NIR light, flush it and then expose. This fills the traps prior to the exposure, placing the sensor into a known state. During exposure some of the charge leaks from the traps into the image. The charge leakage will leave patterns observable in the image that are a form of Dark Fixed Pattern Noise and these are removed completely by proper dark subtraction. However the average signal level leaking from the traps does contribute to the dark shot noise (typically dominates versus thermal dark signal) so this needs to be considered along with the thermally generated dark signal when selecting a cooling target operating point.
Surprisingly, the KAF series require substantial cooling in order to support longish exposures and low read noise target limits (pages 11-19)
Hope that was helpful.