What, exactly, used to be so difficult about this? Do new satellites take time to train their orbits, or test systems, or otherwise commission when they're in the air? What technological changes have happened recently to make this feasible?
Maybe it wasn't always "so difficult", but was never worth optimising for before. I don't know of anything that would have made this impossible before if it had been made a design goal.
Pretty much. However the biggest change is the capability per kilogram that can be packed into a satellite now. Today a 3U cubesat weighing roughly 20kg can take thousands of images, process them locally to manage for weather effects and orbital skew, and then transmit them to a single ground station several times a day.
The first spy satellites weighed tons and literally dropped film canisters back into space to re-enter and be recovered by aircraft or helicopters which would then take the film to a laboratory to be developed[1]. Typically, a satellite would take a few weeks to both stabilize in the environment in space, calibrate its optics, take some test photos for analysis, re-calibrate based on those results, and then be ready to go.
It's also a bit hard to gauge and appreciate how much of an improvement the feat in TFA is, without previous numbers to compare to (preferably in the form of a timeline).
