RLCR methodology: scope control improvements to reduce unnecessary rounds

[new-TonyWang]

Context

After completing a 14-round RLCR session, a methodology analysis identified that the session was roughly twice as long as necessary due to scope control gaps. The core loop (implement → review → iterate) works well, but the absence of mechanisms to bound validation depth and freeze review expectations led to diminishing returns.

Observations

1. Incremental sub-case discovery: The reviewer discovered one more validation sub-case per round instead of enumerating all related checks upfront, extending the session by ~5 rounds.
2. No diminishing-returns detection: 90% of value was delivered in the first 5 rounds; the remaining 9 rounds addressed progressively narrower edge cases.
3. Over-tightening regression: Late-round validation hardening rejected a valid edge case, requiring additional rounds to relax it.
4. Visibility gap: The implementer claimed completion in 11 consecutive rounds, each rejected — suggesting misalignment between implementer expectations and reviewer standards.
5. No lessons captured after round 0: The lesson-learning mechanism didn't capture process-level insights from the refinement phase.

Suggestions

1. Exhaustive sub-case enumeration: When opening a validation gap, require the reviewer to list all known sub-cases upfront rather than revealing them incrementally.
2. Validation contract freeze: After the first round that opens a gap, the reviewer commits to a bounded checklist. New sub-cases of the same category are acknowledged as reviewer oversights, not new blockers.
3. Correctness vs. robustness distinction: Correctness gaps (wrong output for real inputs) = P1. Robustness gaps (rejecting malformed internal data) = P2/P3, batched into dedicated rounds.
4. Round budget with diminishing-returns check: After 2x the estimated round count, require a cost-benefit assessment before continuing.
5. Machine-readable acceptance checklists: Replace prose required-implementation sections with structured, independently-testable checklist items.
6. Explicit plan evolution for scope expansion: Require formal plan amendments when review findings exceed original scope.
7. Periodic goal alignment checks: Lightweight drift detection every 3-4 rounds.
8. Process-level lessons: Expand the lesson mechanism to capture methodology patterns, not just technical discoveries.

Impact

These changes would primarily reduce round count for sessions where validation refinement dominates the later rounds, without compromising review quality for the high-value early rounds.