It's relatively easy to do the mass replace in master. You leave the burden of the replacement in the other branches to people who own those branches.
Set up an automated script, to keep the 'unreplaced' version out of master (and people can also use that script to check their own branches).
Not sure if that counts as non-trivial, yet?
There are probably some other trade-offs involved. The people running kernel development ain't idiots.
Assuming perfect tooling and infrastructure doesn't sound like it necessarily have to be a big deal.
But, how far into fantasy land are we? How much effort would be required to get there? Is this and all similar problems large enough to justify it?
Is adequately tooling and infrastructure even possible?
Does that make proposed changes, such as moving to C11, order of magnitudes more difficult?
The only major issues is that the change is let half-way. But that is only possible with MAJOR undisicpline.
---
The major point, to me, is that this EXCESSIVE fear of breakage is not good. Yes, making upgrades is pain, but you CAN make the pain tolerable with some planning.
Refactoring is like exercise for code: Everyone dislike the idea of excessive, but is GOOD.
And lets be honest, among all the things that could requiere a refactoring, this case is on the most simples of the simples scenario..