Applying feedback_register_revert_resurrection to its own discovery:
auditing every row in a8a7d64b's register hunk found the inverse defect
in the same diff. The revert resurrected retired AD-55 AND deleted
then-active AD-56 (the plumb-fall freeze adaptation, split out of TS-4
by the commit being reverted). Nothing ever restored it, and its
condition came back to life when TS-4's Path-6 shortcut deletion
re-landed for real at Slice 2B — so for eight days a live, test-pinned,
deliberately-documented behaviour had no register row, which is 'a bug
twice over' by the register's own rule.
Restored verbatim from 5e2be19b with a provenance header. One revert,
two silent register corruptions in opposite directions — the memory
rule's audit step is not optional. AD section 50 -> 51.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
5a317460c5