Enabling the next iteration of the borrow checker on nightly

Rust’s borrow checker is gaining flow sensitivity. Today’s NLL implementation treats every borrow as live for the entire scope implied by its lifetime annotation. Polonius Alpha instead tracks liveness per control-flow path, so a borrow returned in one branch need not block mutation in another. The practical consequence is that idiomatic patterns like `get_mut_or_default` compile without requiring `unsafe` or manual refactoring. Engineers can expect fewer false positives, but the new analysis is not a strict superset: some programs that compiled under the original Polonius formulation will now be rejected.

Performance regressions are the primary adoption cost. The Rust team measured the top 10 000 crates and found most regressions below 1 %, but a handful of borrow-heavy crates slow down 2 to 3×. These regressions stem from the additional dataflow facts Polonius must compute; the team is still triaging whether optimizations can claw back some of the lost time. Teams should run their own benchmarks on nightly to quantify the impact before stabilization, especially if they rely on many nested borrows or large data structures.

The feature is currently opt-in on nightly, so engineers can test it without disrupting production builds. Enabling it requires a single flag in `rustc` or `Cargo.toml`. Diagnostics are not yet finalized, so error messages may differ from what NLL produces. The Rust team is soliciting feedback on performance, soundness, and diagnostics; any issues should be reported on GitHub or Zulip. Stabilization is planned once regressions are addressed and diagnostics are polished, likely within the next few months.