🔍
Home · Articles · Lightning
Empirical · Lightning · 2026

We logged 2.4M routing attempts — here is what broke

Synthetic benchmarks lie. This write-up mirrors the forum thread: a routing node instrumented for attempt-level telemetry, bucketed by amount, fee, and channel depth. Figures are illustrative pedagogy for how LN fails in practice.

On-chainLN

Method sketch

Record each pathfinding attempt: destination, amount, proposed fee, failure code. Aggregate weekly to remove outlier hours. Compare small vs large payments — large payments fail more often without deep liquidity.

Representative outcomes (educational, not a guarantee)

In stressed network conditions, success rates can cluster in the ~70–85% band for non-trivial amounts unless you aggressively overpay fees or pre-allocate liquidity along known paths. Your mileage depends on topology.

Operator takeaways

Reproduce, don’t trust: run your own node and publish methodology.
Educational content only. Not financial advice.