Lessons learned from 11,000 agent runs that went sideways

Eighteen months of durable multi-agent workflows, described the way I would want it described to me: numbers, failures, then opinions.

The short version

duplicate side-effects fell to zero once every tool call became idempotent and checkpointed

What we tried

Move	What happened
Bigger context window	Cost up 3x, accuracy flat
Prompt rewrite on its own	+4% on the golden set, then regressed two weeks later
Swapping LangGraph for a boring alternative	Same quality, fewer moving parts

What broke

in-memory agent state is a lie the moment a worker gets rescheduled

Behind almost every incident was a default we inherited and never questioned. A token limit too small. A timeout copied from an HTTP tutorial. A retry loop with no ceiling.

By the numbers

• p50 looked fine the entire time; p99 is where users actually lived
• 60% of "the model is wrong" tickets were retrieval returning the wrong document
• Rolling back a prompt was faster than fixing one — every single time

What I would do differently

1. Instrument first: inputs, outputs, latency, token counts, on day one
2. Build a 200-example golden set before the second feature, not after the fifth
3. Make every external call idempotent and deadline-aware
4. Read my own traces for an hour a week, on the calendar

The uncomfortable lesson

Most of what we shipped was ordinary backend engineering wearing an AI hat. Copy it, change the limits, put the verification in CI, move on.