Most time: the field-path case, bag.items.drain() on a let mut struct. My first fix (treat any field path as emptied in place) was too broad. It rewrote node_numbers.rs into rustc-invalid Rust, and that broke the build. I couldn’t regenerate with the broken binary, so I had to git checkout that one generated file, rebuild and run lower-std again. The other cost was ledger_items_stored: its parameter is lowered as &[..], so mut codes plus codes.drain() was needed, and then the Rust caller had to pass &mut.
drain lives in check_core_call_on.nika (~:984), and the map_drains and kept_drains facts sit beside it.emit_core_calls.nika :981.into_iter() and moves the field.node_numbers.nika depend on that partial-move behaviour.ledger_items_stored must take mut codes to drain it, and the contracts/mod.rs caller has to change with it.--both-settings.contracts:: lib tests, it throws ledger contract (144 tests), clippy and pre-push.lower-std cycles around the bad node_numbers output. The ledger_items mismatch could have been seen from one lower-std before building.builder.md. It would say that when a checker change alters how drain is lowered, I should run lower-std and diff lowered/ for files I didn’t touch, such as node_numbers.rs, before building. It would also say that a broken lowered file is restored by name with git checkout, since the broken binary can’t regenerate it. The brief should also state that a field path of a local is not covered by the existing kept-drain logic.