You need 51% of the network's hashrate to manipulate the block chain. A miner with 51% of the hashrate could already manipulate the transaction fees.
No miners would intentionally re-org blocks, that is quite stupid. With your proposals, certain miners can influence the size of the blocks for the next epoch for whatever reason they'd like. But the point of my post is more about why should we use a dynamic block size instead of a fixed block size increment? It doesn't necessarily bring any benefits either as using a larger sample size only serves to delay the block size change... It would be quite possible that the larger block size wouldn't be required after the block size changes.