Changes touching this path

• a tree write from the store no longer forgets the paths it skipped for want of a key, so a key filed after it no longer turns them into deletions: materialize_tree, materialize_target and the surface each cleared the never-materialized record whole, and a path they could not open was neither written nor recorded, so once loot pull-grants filed its key, or a lane hold refreshed, the next capture found a base path it could open, absent from disk and unrecorded, and recorded its deletion, which twice captured web-written ticket files a land had just folded in. each now keeps the paths of its tree it did not write, inside the view and absent from disk, read off the tree only when the walk counted a skipped path, so a write that skipped nothing reads nothing more. pinned: a_path_a_tree_write_skipped_for_want_of_a_key_is_not_a_deletion_once_the_key_arrives over the surface, materialize_tree and materialize_target, whose surface arm read the skipped path as Deleted at status before the fix, and a_skipped_path_the_disk_holds_is_not_recorded_as_never_materialized. red under named mutations, each restored, 1 failed and 1 passed over the two tests each time: the settle clearing whole as before, materialize_tree keeping nothing, the surface keeping nothing, materialize_target keeping nothing, and no absent-from-disk filter. cargo test green, 4878 passed over 145 test result lines with 15 ignored. CONTEXT.md says so, and the verb tier width is 423 methods. owes no forge deploy; a position holding a skipped path unrecorded records it at its next surface or materialize. it answers the first defect of kmsukvmw and leaves its ferry_remedy half (kzknpprm) 6e5a4aa2 · 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.