Changes touching this path

• the ferry test named for reconcile_merge is renamed to the_adr_0039_carry_does_not_resurrect_files_deleted_on_the_spine, because a probe on the first line of each seam says it enters reconcile_carry then carry_line then CarryOutcome::Carried and never enters reconcile_merge or merge_tips - the name and the inline comment about the ordinary reconcile-merge path have been stale since ADR 0039 moved a described self-authored line onto the carry. the sibling sweep was done the same way rather than by reading: 16 named seams probed across ferry.rs and reconcile.rs plus converge, converge_heads, merge_tips, carry_line, seal_wip_gate, finalize_merge, fast_forward_to and refuse_open_conflicts, with each of the 53 ferry tests run alone so the probe output is attributed to one test, and the headline test is the ONLY misnamed one. a_keyless_reconcile_merge_writes_the_merged_tree_to_disk really does enter reconcile_merge and is the only ferry test that does; every ingest-named test enters ingest_commit, every adopt-named one enters reconcile_adopt, and tag, rollback_note, project_wip and reconcile_capture all enter what their names promise. the two that reading made me suspect - amend_of_landed_threads_onto_predecessor_as_a_fast_forward and bootstrap_adopts_prebridge_history_without_ingesting - are naming the git-side shape and the baseline note, not fast_forward_to and reconcile_adopt, so they stay. the ticket got one thing wrong and it is the larger finding: it asked me to confirm the pin still fails for its original reason, and it does not. restoring the pre-9763144 union-the-ancestry tree_at leaves this test and all 53 in the module GREEN while the two loot-core pins go red on the same mutation, which is the positive control saying the mutation was live. the cause is structural: carry_line replays the full manifest each node records against a tree_at base, so an absent path reads as deleted-since-the-fork and the deletion wins even when both trees are polluted, where merge_tips took tree_at on BOTH sides and read the stale entry as untouched. so this is a live pin on the composition the carry performs and #288 is guarded a layer down, not here - recorded in the comment rather than fixed, since a ferry-layer #288 pin is a separate ticket. no behaviour changed: ferry.rs is the only file touched and only comments and one fn name in it (#1479) 093b8a20 · dbf3dbe6…

Renames are not followed. loot's tree maps a path to an address, so a rename is a delete and an add. This list is the history of the name, not of the bytes.