[joshrotenberg]

Append general memory interpretation guidance to redis_health_check and redis_memory_stats: consider absolute byte overhead, allocator metrics and prior peaks, especially for small current allocation. The note explains that a high fragmentation ratio alone does not establish a memory problem. Existing values, fallbacks and raw nested or unknown metrics stay intact. No cutoff, diagnosis, recalculation or extra Redis requests are added.

This is a bounded part of #1218 and leaves #1218 open. FT.PROFILE formatting is separate.

Validation uses actual-tool synthetic RESP2/RESP3 fixtures, with exact command sequences and error boundaries. The same corpus against the original handlers passed the error test and failed the three note tests after verifying all 14 existing output prefixes. It now passes all four tests with minimal database/test-support features and all features. The catalog contract, formatting and strict all-target/all-feature Clippy pass. No live memory-pressure claim is made.

Independent source review found no actionable findings at 76cd4c0. Hosted unit tests, integration tests and Linux/macOS/Windows builds pass at that head. Security Audit and advisory checks fail on the unchanged lockfile; the bans/licenses/sources sibling job was cancelled.

Cargo.lock is unchanged. Keep this draft for the separate #1335 security update, branch refresh and current-head checks/review after that refresh.