---
name: lead-with-the-defect-that-needs-no-box
description: "When filing a flake/defect, lead with the statically-provable part and file the observed rate separately — evidence you must discount makes the finding look weaker than it is."
metadata: 
  node_type: memory
  type: feedback
  originSessionId: 82118e6c-5d10-4230-aef4-ed67f20814ba
  modified: 2026-07-20T09:29:57.126Z
---

Filing rule adopted by doyle 2026-07-20 after todlando's correction, generalized to the whole load-window flake family: **separate the statically-provable DEFECT from the statistically-estimated RATE, and lead with the one that needs no box.**

Worked example: a real `Broker::bind` + `serve()` unit running in nextest's LIGHT pool is provable from `.config/nextest.toml` alone (two overrides, neither matching `package(spt-daemon) & kind(lib)`) — no timing, no sighting, no box. The 240s timeout that surfaced it was observed during a CI window on the same machine and had to be discounted for contention; the v0.32.0 CI sighting carries the rate.

**Why:** doyle led the seed with the discounted observation and hung the argument on it. An honest caveat then did work the argument never needed and made a statically-certain finding read as shaky. Overclaiming and under-claiming are the same defect — a claim keyed on the wrong evidence.

**How to apply:** before filing, ask which half of the finding survives with zero runtime evidence. That half is the lede. Timing/frequency goes in its own paragraph, labelled as rate, with its contamination stated. Related: [[ground-dont-assume-on-incidents]] · [[measure-the-box-before-the-instrument]] · [[gate-rig-mechanisms-not-remembered-steps]] · [[teardown-authority-progress]]