use_small_heuristics = "Max"
is confirmed to work great on my code and reduce churn.
Here are some stats from formatting changes:
use_small_heuristics = "Default"
src/librustc_privacy/lib.rs | 574 ++++--
src/librustc_resolve/build_reduced_graph.rs | 576 ++++--
src/librustc_resolve/check_unused.rs | 65 +-
src/librustc_resolve/error_reporting.rs | 46 +-
src/librustc_resolve/lib.rs | 2756 ++++++++++++++++-----------
src/librustc_resolve/macros.rs | 646 +++++--
src/librustc_resolve/resolve_imports.rs | 1008 ++++++----
use_small_heuristics = "Max"
src/librustc_privacy/lib.rs | 466 ++---
src/librustc_resolve/build_reduced_graph.rs | 463 +++--
src/librustc_resolve/check_unused.rs | 55 +-
src/librustc_resolve/error_reporting.rs | 23 +-
src/librustc_resolve/lib.rs | 2426 +++++++++++++++------------
src/librustc_resolve/macros.rs | 555 +++---
src/librustc_resolve/resolve_imports.rs | 913 ++++++----
The remaining churn is mostly the switch from visual indentation to block indentation, which is okay and expected, and is indeed a consensus of the style RFC process.