[shrsv]

Drops the pre-allocated capacity hint on Keys() for simpler code.

[livereviewbot]

LiveReview automated review

Scanned 1 changed file (+9/-3).

🔴 Critical — Performance (internal/core.go:33): the inner re-validation loop runs O(n) work per key, turning Keys() from O(n) into O(n²) -- and it does nothing (s.data[k] cannot change mid-iteration while the read lock is held, so the check is always true). On a store with a few thousand keys this turns a cheap call into a measurable stall; it's also entirely dead code relative to the comment's stated intent.

[livereviewbot]

Auto-approve policy evaluation

Stage	Result
Eligibility	✅ pass
LLM classification (refactor-only, confidence 9)	✅ pass
Findings	1x critical / performance
Risk score vs. threshold	❌ fails -- same mechanism as the security scenario, different taxonomy category

Verdict: BLOCKED at the risk stage. This is the companion to critical-finding-in-security-blocks: the category-weight path isn't special-cased to security. A critical finding in any sufficiently-weighted taxonomy category -- performance included -- blocks on its own.

[livereviewbot]

Auto-approve policy verdict: BLOCKED (risk stage). One critical/performance finding blocks, same mechanism as the security-category case.