BIGINT UNSIGNED AUTO_INCREMENT PK

UUID in secondary column

UUID stored as BINARY(16)

utf8mb4 character set

utf8mb4_0900_ai_ci collation

DATETIME not TIMESTAMP

NOT NULL preference

Lookup tables over ENUM

DECIMAL for money

Appropriate VARCHAR sizing

3NF normalization

Equality before range in composite index

Range stops subsequent index usage

Remove function on indexed column

Cursor pagination for Query C

No SELECT * in rewritten queries

UNION ALL for Query D

EXPLAIN red flag identification

Composite index for tenant + status queries

Index for GROUP BY optimization

Separate indexes or UNION for OR across columns

DESC in index or ORDER BY awareness

Consistent row access order

Error 1213 retry with backoff

I/O outside transaction

REPEATABLE READ as default

FOR UPDATE used sparingly

Short transaction duration

Backoff uses exponential delay

Max retry limit

Idempotency consideration

No global isolation change

Indexed WHERE columns

RANGE partitioning used

Partition column in PRIMARY KEY

Partition column in UNIQUE KEY

MAXVALUE catch-all partition

DROP PARTITION for retention

REORGANIZE for new partitions

Monthly partition boundaries

BIGINT UNSIGNED AUTO_INCREMENT PK

No foreign keys on partitioned table

DATETIME not TIMESTAMP for recorded_at

Composite index for device lookups

INSTANT for adding column

INPLACE for adding index

Explicit ALGORITHM specification

Column type change risks identified

External tool for large table rebuild

Batch update sizing

Connection pool formula

LOCK=NONE specified

Rollback steps included

Backfill uses batched WHERE with range or LIMIT

Replica testing recommended

N+1 fixed with batch/join

NOT EXISTS instead of NOT IN

SELECT only needed columns

Covering index designed

Filter columns first in covering index

Batch IN list sizing

performance_schema N+1 detection query

Avoiding SELECT * rationale

InnoDB implicit PK in secondary index

Batch consolidation over sequential per-row queries

Join vs IN trade-off addressed

performance_schema for unused indexes

Write-only indexes flagged separately

Redundant index detection

Invisible indexes before drop

Counter reset warning

Prefix indexes for long strings

1-5 indexes per table guideline

EXPLAIN verification

DROP uses ALGORITHM=INPLACE

UNIQUE/constraint index exemption

Write amplification explanation

PlanetScale recommended

Stale read problem explained

Post-write reads routed to primary

GTID-based lag strategy included

Seconds_Behind_Source limitation noted

Connection pool formula applied

Separate pools for primary and replica

Large transaction replication warning

Replica monitoring alerting

DDL replication impact noted

Write to primary only

Assume replicas always slightly behind