Advisory
Rate limiting follows the credential, not the address
Why a limit follows the thing being guessed rather than the address it came from, and how RJV applies it.
Limiting by address punishes everyone behind one office or VPN address and does nothing against an attacker who has many addresses. A limit is more useful when it follows what is being attacked: the account handle, the key or the registration route.
What RJV does
Every authentication attempt is recorded in PostgreSQL against a subject key, which is stored only as a SHA-256 digest. Before a sign-in, a password change or a key use is processed, the platform counts that subject's attempts: it refuses once the subject has made 30 attempts in five minutes or five failed attempts in fifteen. Registration, which every new account shares, is limited by volume alone (300 in five minutes), so one person's failed sign-ups cannot close registration for everyone. Directory provisioning has its own budget.
What it does not do
The application holds no notion of the client's address, so it applies no per-address limit. A limit by address, if wanted, belongs at the edge in front of the application. The counters live in the database, so they survive a restart and are the same on every instance.
See the Identity and Access page for the sign-in controls this limit protects.