I am no expert on sattelite. But have seen these modern sattelite images. It is clear to me that they can see and distinguish ground target very well. Even well camouflaged or mobile ones. Knowing what sattelite is passing or not may be useful, only the enemy knows this also, and they will attack at this time, knowing ground target are hidden or moving away. Russia or China have information about sattelite and they can target with weapon. India may be able to do it too. But Iran has no such ability.
Agree about what you mention about costs. Different ways to hide. Dummy sites or buried sites. AAA versus missiles. But we are not counting costs in terms of lives lost or territory lost. It is not gradual calculation. It is either RADAR or no RADAR. we can not have SOME radar. Or part of a radar. Therefore probability of being hit must be zero. Or near zero. Otherwise all the expense and effort useless. So scientist must work out exactly a way, when NO observation of RADAR site possible . Or has very low probability of observation and being hit. All you need to destroy a radar, is ONE lousy drone out of many hundreds that get through.
Agree that underestimating opponent is not good idea. The yank imagined that Global hawk was good. Based on previous success against inferior enemy. But it was hit at 60,000 feet by Iran. So we must be ready and predict what capability they have. One of these cheap drones has four hours flight time. So making one with six or ten hour flight time possible. At 100 km / hr, then range 1000 km.
Hitting a SAM vehicle with several ( more than number of missiles carried) drones simultaneously, will destroy it. Even if well manned and armed. No need to have vehicle. Just single buried missile tubes, dispersed.
I would say the same about radar. A vast radar but dispersed over land. Consisting of nodes. Like taking radar dish and chopping it up into hundreds of pieces. Then these nodes work together. If a node is hit, it is cheap. And radar works well still. So to disable radar becomes near impossible. Neural net or human brain model.