Secret retrieval keeps up at scale

We reworked the runtime retrieval path so it serves many simultaneous reads in parallel and does less work per read. When a fleet of machines or a burst of workloads pulls secrets at the same moment, the service now handles far more of them at once without them slowing each other down.

A slow database stays contained

Database work is now isolated, so a single slow query can no longer tie up resources that other requests need. The platform stays responsive even when the database is under load, including health checks and routes that never touch your data.

Tighter brute-force protection

Lockouts after failed machine and AI agent authentication now count precisely, even under a burst of simultaneous attempts, so the protection holds firm under concurrent pressure.

Steadier over long uptimes

Internal rate-limit and lockout bookkeeping now prunes on a schedule, keeping the platform stable over time.

Nothing to change on your end

Your SDKs, CLI, and API work exactly as before. These improvements apply automatically.