In addition, a core Zig developer has explained why the PR was rejected, because it would introduce non-deterministic bugs into the compiler, just to achieve a speedup Zig is already gaining thanks to recent work on the self-hosted backend and incremental compilation, which are far more general as well.
As often happens, the online discourse has, for some reason, decided that this was an anti-AI stance, while - as far as I understand - the problem was simply that the PR had problems, which lead to Bun forking Zig.