Maybe, I shouldn't reply to myself, but I feel like I have oversimplified. First, not all observatories use guiding cameras, but most do. An obvious exception a

Maybe, I shouldn't reply to myself, but I feel like I have oversimplified. First, not all observatories use guiding cameras, but most do. An obvious exception are radio telescopes which do passive tracking, but will check their pointing on a bright source periodically. Second, the sky is very bright and variable in the infrared, so the dominant source of noise is from the sky. For this reason, you take an image on source followed shortly by an image off source to use to subtract the background sky from the on-source image. As the infrared sky can vary quite quickly, you want to sample the background quite often. If there is a "blank" patch of sky near enough to your target, this is accomplished by "chopping" the secondary mirror back and forth between the source and the off position. This occurs at frequencies on the order of 1Hz, so exposures are less than a second. If you have to move further to find "blank" sky, it is accomplished by "nodding" the telescope, i.e. moving the primary to the off position. This takes a bit longer and may only be done every ~30s. Because there is active guiding, there is no issue co-adding an arbitrarily large number of these images, and observations can actually be several hours long. And, because the background sky varies fairly rapidly, stacking short exposures (with the background subtracted) will actually give better noise performance than taking long exposures.