No one knows, and I don't think there's been (a public) discussion about it by projects leads. The only guarantee is that 1.x will not break anything (according to some definition of "break").
There needs to be more experience with widespread use and multiple releases, to see how far that backcompat can be held. Perhaps after 10 releases, it should be a lot more clear whether or not a Rust 2.0 will be needed.
Thanks; I'm not sure in that case saying 1.x is guaranteed to not break things is very useful at all. Systemd (as far as I know) follows semantic versioning, but their version number is up in the hundreds. Without some sort of guess at how long a release might last, 2.0 might as well ship the minute after 1.0 ships. (Okay, probably not...)
The core team has reserved the right to break backcompat without changing the major version if they find a serious problem in 1.x series. So I expect that it will be a while (at least a few years) before 2.0 is created.
There needs to be more experience with widespread use and multiple releases, to see how far that backcompat can be held. Perhaps after 10 releases, it should be a lot more clear whether or not a Rust 2.0 will be needed.