Darius Tran

Great point. We spend a lot of time evaluating retrieval quality, but not enough time testing whether retrieved documents should coexist in the same context window. Relevance and compatibility aren't always the same thing.

The point that a longer expiry only delays the failure is the part people skip. The stale writer has no idea it lost the lease, so only the destination can refuse it. The same thing shows up with agents that retry tool calls. A paused or timed-out run wakes up and writes after a newer run already did. For side effects the database can't fence, like an email or a payment call, do you lean on idempotency keys with the provider, or an outbox so only the current owner ever sends?

The gap between a good demo and real work usually comes down to the test set. Demos use clean inputs, real work has the messy ones. One thing that helped me: build the eval set from cases that actually failed or got escalated, rather than from examples someone invented. Do you run evals offline before launch, or in shadow mode on live traffic first?