It's worth noting that while the reference has seen minor, but consistent, improvements, there are on-going specification efforts in for example:
- The rustc-guide
- The unsafe-code-guidelines
- The grammar working group (wg-grammar)
- Research work in e.g. Oxide (e.g. @awe et. al's work) as well as in the Rust Belt research.
- wg-polonius and chalk involve spec work as well.
My hypothesis is that one reason the reference isn't in a better state right now is that the foundational work of specifying e.g. the abstract machine, and other parts of the language have not been done. When that's the case, we don't have anything to build on and so some stabilizations are just filed as issues.
While the on-going spec efforts aforementioned might not bear immediate fruits, I think they will eventually improve things a lot.