S3 to Lambda event ordering across object keys: is there a guarantee?

Sequencer field – useful, but not global

The sequencer value that appears in every S3 event record is a 128‑bit opaque token that S3 guarantees will increase for the same object key as successive events are generated. It is not a monotonically increasing counter across the entire bucket or across different keys. Consequently, you can use it to detect out‑of‑order arrivals for a single object, but you cannot rely on it to order events that target distinct keys.

What this means for Lambda

• When S3 pushes events directly to Lambda, the service may deliver them in the order they were generated for a given key, but Lambda’s concurrency limits, throttling, and the internal fan‑out of S3 workers can still cause cross‑key reordering.
• Even if you preserve the sequencer values in your Lambda logs, you will still see gaps or out‑of‑order patterns when uploads to different keys happen in quick succession.

Practical mitigation

• Route S3 events to an SQS FIFO queue (or an EventBridge rule that forwards to a FIFO queue). Messages that share a MessageGroupId – for example, the object key – are processed in strict FIFO order.
• If you need to keep the direct Lambda path, embed an explicit timestamp or sequence number in the object’s metadata and reorder the events inside your handler before any downstream work.
• Reserve Lambda concurrency or use Step Functions to serialize processing once the FIFO queue has delivered the messages.

There are no region‑ or account‑level switches that change this fundamental behavior: S3’s event notification system is intentionally best‑effort and does not provide global ordering guarantees.