But now it doesn't mean that. It means 20 years.
There are extremely good reasons not to increment the major version. For instance tools which take the conservative approach of checking for 2.6 will all be broken. Those tools correctly elected not to risk wrong behavior resulting from a major version change.
When you add together the time it takes to compensate for this change for all maintainers everywhere (tools developers, distro developers, everyone), surely it will run into the hundreds of hours, perhaps thousands. All for, functionally, nothing. And sorry for calling you Shirley.
Seems to be an objectively bad move. Oh well. This needless-on-purpose change will cause resentment, but after time it will dissipate.
What tools actually do this?
In practice, due to the vast number of configuration options, it's not entirely binary stable even across different people compiling the same code. With work, you can probably make things compatible, but it doesn't come for free.