You may be assuming too much centralization about these decisions. Outside the core team, I think it’s pretty much people doing what they consider important; when these things are done they end up in master, then the next RC.
Release candidates have a more-or-less predetermined schedule, so whatever is ready by that ends up in the next version. Beyond this, I don’t think anyone can predict if something will be in the next version.
As an example, consider the recent #33864: it is something that many people consider a significant change, yet just before the 1.4 feature freeze it was suggested that it ends up in 1.5 instead.
You could approach this problem from the other direction: if you have an idea in mind to improve compilation, just bring it up in a relevant issue (eg one like #33326), open a new one, or make a PR. I am under the impression that instead of duplicating effort, there is a shortage of people who can help out with Julia internals, so I believe that offers to help are very much appreciated.