F.3:9 - Self-checks

• Same-conclusion test. Would the candidate uses ever change a source-grounded conclusion relevant here? If not, merge.
• Argument probe. Substitute the participants from one use into the other. If the source proposition fails, split.
• Entailment probe. Does one use imply an occurrence while the other implies a description, kind, or status? Split.
• Label inversion. Read the Plain label alone. If it invites a broader claim, tighten it.
• Counterexample ping. State a short use that must be excluded. If it falls inside the claim, refine the boundary.
• Memory rule. If a careful reader cannot recall the few relevant claims, the partition is probably too fine.