Post

This morning we shipped a fix to Prisma Postgres connection routing to address an increased p99 network latency that sometimes resulted in timeouts on p99.9

This had been slowly creeping up over the last few weeks disguised as intermittent network inconsistencies. We've found the root cause, fixed it, and we're improving our observability to spot and correct these trends much earlier

One thing we added is a daily agent workflow to analyze telemetry, generate a human-friendly report, and propose improvements with comprehensive prompts. It also verifies post-fix This will soon mature to dispatching coding agents to create the PRs autonomously