But, and correct me if I'm wrong, there is't much damping in the correction.
This means that if 50% of miners pull out, then it'll be longer before the correction.
The correction will then "over-correct" in a sense and the difficulty will be easier, and then there'll be a short half-week if that causes all the dormant miners to come back online.
If miners are serious about turning off miners when the difficulty is high we could see a ping-pong effect where we end up with "easy weeks" lasting 2-3 days when the difficulty makes it super profitable and then hard-weeks lasting 2-3 weeks where the difficulty makes it unprofitable.
That will get worse and worse as people are less inclined to mine the hard weeks.
That said, that assumes that the depreciation of miners and overhead of switching based on difficulty makes it worth turning them off rather than just running them full time anyway because there's a time window before they're out-scaled.
( Is mining tech still getting out-scaled or has it plateau'd per device now? Originally when I supposed this theory, mining devices only had a 3-6month window before more efficient devices would replace them) .