EXPERIMENT 57 VERIFICATION
==========================
Compilation tested successfully.
Execution tested successfully.

Replay seed:
9783790101901

Scenarios: 30
Dataset modes per scenario: 2
Total executions: 60
Compile errors: 0
Runtime errors: 0

This is a diagnostic replay of the Experiment 51 holdout.
It adds no independent paired evidence.

Experiment 56 reference pathway counts reproduced:
- P0_UNPERTURBED_COMPLETE_STABILITY: 8
- P1_PERTURBED_COMPLETE_STABILITY: 12
- P2_RESIDUAL_TOPOLOGY_PRESERVED: 7
- P4_SUCCESSFUL_TOPOLOGY_RELEASE: 3
- aggregate passenger delta: +24

Experiment 57 evidence-ablation findings:

REQUIRED
- DIFFERING_POSITIONS
  unique assignments after withholding: 10/30
  ambiguous assignments: 20/30
- PASSENGER_DELTA
  unique assignments after withholding: 0/30
  ambiguous assignments: 30/30
- RESIDUAL_STALL_FAMILY
  unique assignments after withholding: 0/30
  ambiguous assignments: 30/30

SUPPORTING
- DEPENDENCY_DEPTH
- RESIDUAL_TOPOLOGY_TRANSITION
- COMPLETE_CABIN_CONVERSION
- PATHWAY_INTEGRITY_RULES

OPTIONAL FOR PATHWAY ASSIGNMENT IN THIS HOLDOUT
- PERTURBATION_FAMILY
- FIRST_DIVERGENCE
- CRITICAL_TREE_SIZE
- CRITICAL_TREE_DEPTH
- YIELD_DEPENDENCY_TYPE
- FINAL_PREREQUISITE
- REQUIRED_BLOCKER_COUNT

Aggregate evidence classification:
- required categories: 3
- supporting categories: 4
- optional categories: 7

Audit checks:
- all baseline assignments retained: true
- movement architecture frozen: true

Interpretation:
Experiment 57 identifies the minimum evidence needed to preserve unique
Experiment 56 pathway assignment in this holdout. Passenger-delta sign,
admission-order difference status and residual-stall-family identity are
indispensable. Topology transition, COMPLETE_CABIN conversion, dependency
collapse and integrity rules remain scientifically important corroborating
evidence, but are deterministically derivable or validation-oriented once the
required evidence is retained. Other structural diagnostics enrich mechanism
interpretation but do not determine pathway membership.

These are local verification findings. Official findings require the user's
saved run.
