What does it mean that validated schemas still have limits?

Even a well-tested schema may fail in new contexts or at different scales. Validation tells you where a schema works, not that it works everywhere. The boundaries of your tested conditions are the boundaries of your warranted confidence.

Example: A manager develops a schema: "giving people autonomy increases their productivity." She validates it across three teams, measuring output before and after removing micromanagement. The data confirms the schema — productivity rises in all three cases. She applies the schema to a fourth team, a newly formed group with no established norms, and productivity collapses. The schema was validated within its domain — experienced teams with existing competence — but its boundary conditions were never tested. The validation was real. The limits were also real. They were not contradictions. They were the difference between "this works here" and "this works everywhere."

Try this: Select a schema you consider well-validated — something you have tested and believe to be true. Write it down as a single declarative statement. Then systematically probe its boundary conditions by answering six questions: (1) In what specific contexts have I actually tested this? (2) What contexts have I never tested it in? (3) At what scale have I observed it working? What happens at 10x or 0.1x that scale? (4) What population or situation type was present during my validation? (5) What would have to be true about a new situation for this schema to fail? (6) If I stated this schema to someone in a very different field or life situation, what objections would they raise? Write your answers. The gap between questions 1 and 2 is the gap between your validated domain and your assumed domain. That gap is where the limits live.