Born After The Miss

A handle-lookup API caches not-found responses for 60 seconds. New users report that their own profile URL returns 404 for up to a minute after signup, then starts working. All reads go to a single primary database with no replicas. What is the correct fix?

Stop caching not-found responses for this endpoint, since a negative entry is not something the write path can invalidate

Delete the negative entry for that handle as part of the account creation write, treating it like any other invalidation

Route reads to the primary for a short window after signup so replication lag cannot serve an answer that predates the new row

Shorten the negative TTL to a few seconds so the window where the profile 404s is small enough that users stop reporting it