Hashing power can also be bought and sold, just as easily as humans can be. Either buy the hardware yourself, or consider someone offering the top two pool operators a large sum of money for control of their pools.
A 51% attack does not have to correspond with a doubling in the hashpower. An existing pool or large mining operation can turn, which would not show a change in the total hash power.
A 51% does not necessarily require a re-org beyond 1 block. If I want to double spend, then it requires a large re-org, but if I just want to shut down the network, no re-org is required.
True there is no absolute protection beyond the "good guys" having 51% of hashing power. Still making the blockchain "smarter" reduces the effectiveness of a 51% attack.
A 51% attack to perform a denial of service is utterly stupid use of resources. Eventually it will be defeated and then network transaction will resume.
Using a pool for hashing power (or any public hashing power) to form a chain is easily detectable and likely will lead to a) people stop spending and b) people gaining more hashing power to fight off the attack.
The "nuclear" option for a 51% attack is one that is both massively destructive and undetecable until it happens.
Take private hashing power, build a chain in private, don't publish each block, fill the blocks w/ double spends, then publish the attack chain once it is longer than the "good chain" resulting in economic chaos, crashing bitcoin prices, and loss of confidence.
So you are right no protocol adjustment can make the network 51% "immune" but it doesn't have to be. Bitcoin is already very hard to attack, and economical 51% attacks are likely impossible (cost of attack outweights direct profits from attack). Adding block chain "smarts" would make it even harder to pull off large scale disruptive attacks.