Find every place in my data model where a duplicate record would be a bug, and enforce uniqueness at the database level. Go through each table or collection and identify natural unique fields: user emails, usernames, slugs, external provider IDs (stripe_customer_id, oauth provider IDs), and compound uniqueness like one membership per (user_id, team_id) or one vote per (user_id, post_id). For each, add a unique constraint or unique index. Then check the write paths: anywhere the code does a check-then-insert (look up, see nothing, insert), replace it with insert-and-handle-the-unique-violation, because two requests arriving at once both pass the check and both insert. Before adding constraints, find and tell me about existing duplicates so I can clean them up first; adding the constraint with dirty data will fail. Handle case sensitivity deliberately for emails and usernames (store or compare normalized). Do not change the user-facing error design; map unique violations to friendly messages ('that email is already registered'). Deliver: the constraint list per table, the duplicate report, the write-path fixes, and a verification: fire two identical signups concurrently and confirm one wins and the other gets a clean error instead of creating a second account.