Multi-Region Drift

The Failure Mode

Suppose one request path touches:

• api-sg-1
• worker-us-1
• db-eu-1

Each node emits timestamps from a different machine clock.

If the Singapore node is a little ahead and the US worker is a little behind, naive sorting may tell a story like:

1. email_sent
2. user_created
3. billing_initialized

That timeline looks precise. It may also be nonsense.

Why This Happens

Wall clocks can drift. NTP can lag. Virtual machines can pause. Cloud instances can disagree by just enough to create believable but false event order.

What The Library Should Surface

In a case like this, the library should distinguish between:

• HLC or timestamp-based ordering that is only derived
• explicit dependencies that make some relationships proven
• cross-node events that may need to remain unknown unless stronger evidence exists

This is especially important when a user expects a total order simply because every record has a timestamp.

The Business Value

Without this distinction, a team can produce a confident postmortem timeline that is cleaner than reality.

That is exactly the kind of false certainty causal-order is designed to prevent.