refactor(rlix-validation): remove deprecated no-op assert_partial_overlap_standalone_safe

[TianyeGGBond]

Context

assert_partial_overlap_standalone_safe was originally meant to refuse partial-overlap topologies under the standalone (train_async.py) entry. Its own docstring explains it was reduced to a no-op because intent can't be safely classified from args alone at the standalone entry (the standalone create_placement_groups path offsets the rollout pool, so a valid disjoint config also looks like zero-based train ⊂ infer). The reliable signal is the RLIX_CONTROL_PLANE=rlix env var, and RLix-mode correctness is already enforced by assert_rlix_topology (C1).

def assert_partial_overlap_standalone_safe(args: Any) -> None:
    """Deprecated no-op kept for callsite stability. ..."""
    del args
    return None

Problem

The function does nothing and — verified across the miles tree — has no callers. It's only referenced by its own __all__ entry. Keeping a no-op in the public API is dead weight and misleads readers into thinking a standalone guard exists.

Change

Delete the function and its __all__ entry. No callers, no behavioral change.

🤖 Generated with Claude Code