Except that it never was (emphasis added by me)
Excuse me, let me correct it:
Yes. That was actually the vision what happened at first.
If lets say 90% of miners decide to quit for some uknown reason, the difficulty will not change instantly.
If 90% of the
computational power was stopped being offered to the Bitcoin network, then it depends on the block that it'll happen. It'll surely slow down the process independently of the block, but if it happens 5-10 blocks before the next adjustment, the damage won't be that big. Each of these blocks timestamp would differ around 200 minutes. But if there are thousands of blocks until the next adjustment, there's trouble for sure.