or
pools see it can be a cluster fuck and dont go passed a limit unless majority of preference says its not going to be a clusterfuck
A simple block acceptance size with an acceptance depth of '1' achieves that without the unnecessary complexity.
the blocks are "accepted" EG say a block A 0.251 received and also block B 0.249
both sit there accepted.. but not set in stone.
the acceptable depth is the deterrent.. for pools
pools wont want to risk going to 0.26 unless high majority say its ok. bcause their is a risk their 0.26 block may get rejected if the majoriry dont want anything above 0.25 (yes im using low numbers so people can imagine dynamics in a scenario of todays 1mb consensus and nodes 'preference' below that)
Acceptance size set at 0.25, Block A gets rejected end of. Block B gets built on. Acceptance size moved to 0.3, Block A or B win the block propagation race as exists now an will get built on with blocks of up to 0.3MB in size.
Personally I think it's time to get rid of this protocol enforced block size limits all together. We have SPV wallets now. Miners should be able to create blocks of dynamic sizes which they see fit to maintain the network and manage the network mempool demand with slight fee pressure.
The protocol enforced limit is creating problems that need to be solved. Removing it means it will be working as intended in accordance with the original whitepaper.
There are enough exchanges, online wallets, payment providers, other businesses that can cope with data centre traffic levels to ensure the blockchain database is validated and distributed. Let the miners get on with building the blockchain.