Postgres Summit US2026

A simple INSERT that normally completes in milliseconds suddenly takes minutes. Latency becomes wildly unpredictable, and your monitoring lights up with wait events you've never seen before. The culprit isn't a missing index or a bad query plan — it's something far more subtle hiding deep inside PostgreSQL's storage layer.

This talk tells the story of a real-world performance crisis caused by OID exhaustion in TOAST tables. PostgreSQL's TOAST mechanism silently moves large column values into separate tables, assigning each chunk a unique Object Identifier from a finite global pool. For most workloads this is invisible, but in high-churn environments the OID space can fill up, sending PostgreSQL into an expensive search loop that turns routine writes into system-wide bottlenecks.

We'll cover how to recognize the symptoms, confirm the cause, and understand why certain write patterns accelerate OID consumption in ways that aren't immediately obvious. Then we'll walk through solutions — from emergency triage techniques, to long-term architectural strategies that prevent the problem from recurring as your data grows. You'll leave with practical monitoring approaches to catch OID pressure before it becomes a crisis.