CAIN-42 E38 -- Byzantine Mission Integrity & Long-Horizon Objective Governance (PRE-PRODUCTION library).

AUTONOMY MAY OPTIMIZE THE PATH. GOVERNANCE DEFINES THE BOUNDARY.

  objective.py  Mission Governance Object (MGO), Objective Integrity Root (OIR), objective change classification
  lineage.py    GoalLineageObject: MISSION -> OBJECTIVE -> SUBGOAL -> PLAN -> ACTION with strict attenuation
  ledger.py     MissionLedger: single locked state machine, hash-chained events, two-phase authorize/commit,
                governance proof + non-authorization proof; integrates E35 integrity and E36 collective gate
  outcome.py    outcome proof (no self-declared success), failure proof, continuation requirements
  proxy.py      proxy-objective detector, objective drift engine, Goodhart lab (measured coverage + blind spots)
  conflict.py   MissionConstraintGraph, conflict engine, counterfactual comparison (informs, never authorizes),
                mission memory root
  boundary.py   MCPGate: commit -> pinned-key single-use token -> out-of-process tool (E34 boundary)
  replay.py     mission time machine: recompute state from the event log (REPRODUCED/DIVERGED/INCONCLUSIVE)
  federation.py E37 integration: a remote domain may propose, never redefine the mission
  lab.py        attacks, invariants, races, MVP acceptance

What this is NOT: it is not deployed on the live clusters or wired into the hosted gateway; it does not
solve alignment; the proxy detector only sees independently measured metrics.