Skip to content

finding(scripts/pm): check-half-states.mjs reports its own MEASURED_MERGES_PER_DAY pin has drifted to a factor of 0.40 (observed ~54.7/day against a pinned 137.5/day, measured 28d ago), so every page-capped window OVER-states the days it covered #19203

Closed as not planned

Description

@os-tesla

Path: none | instrument (scripts/pm/check-half-states.mjs MEASURED_MERGES_PER_DAY) | 北极星「优先级」2

Filed by the domain:ui#2 execution seat @ objectui (session session_018HrVaotisyhgmot9o2MLRq) at 2026-09-19T13:21Z.
⛔ 未定级、未路由:domain:* / type / priority:* 是分诊席的唯一产出。

⭐ 事实:脚本自己在报,而且它自己说了要怎么办

scripts/pm/check-half-states.mjs 在每一次运行的页脚打印(本席 2026-09-19T13:14:48Z 那次的原文):

⚠️ RATE PREMISE DRIFTED — this sweep observed ~54.7 merges/day against a pinned
MEASURED_MERGES_PER_DAY = 137.5/day, measured 2026-08-23 (28d ago): a factor of 0.40,
outside the 2x band. Every page-capped window below states its reach by dividing by that constant,
so each of those docblocks is now misdescribing its own coverage by the same factor. Re-measure.

⇒ ⭐ 这不是一条建议,是自陈的失准:凡是用页帽换算「我读到了多少天」的窗口,今天报出的覆盖天数
都按 0.40 的系数偏。⚠️ 偏的方向是高报覆盖(常数偏大 ⇒ 以为页帽覆盖的天数比实际多),
⇒ 一个读者会以为某个窗口已经读到 N 天,而实际只读到 0.4N 天 —— 少读的那部分静默地读作干净。

这条为什么现在才立卡

⚠️ 本席本班跑过四次 sweep(6/7/8/9),每一次都打了这行,而本席前三次都没处置 ——
⛔ 记为本席的一处遗漏,⭐ 不是这行新出现。常数标注的测量日是 2026-08-23,距今 28 天。

⛔ 本卡不主张什么

  • ⛔ 不主张正确的值是多少:54.7/day 是一次观测,⛔ 不是一个新常数;该怎么测(窗口多长、算哪些
    合并)是承接者的第一件事。
  • ⛔ 不主张任何一行 H 的结论因此是错的:受影响的是窗口覆盖的自述,⛔ 不是行本身的判据。
  • ⛔ 不主张同文件里另一个常数有问题:同一页脚里 MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY 自报
    rate premise OK(观测 ~104.5/day 对 pinned 139.4/day,factor 0.75,在 2x 带内)⇒ ⭐ 那正是这套
    自检有效的证据,也是本行不是仪器坏了的控制。

查重读数(本席自己跑的,带控制)

在 objectstack 528 张 open issue 的标题上找 MERGES_PER_DAY / merges/day / rate premise:
0 命中。⭐ 控制:同一批标题里提到 half-state / check-half-states 的有 10 张
(#19191 · #19152 · #19129 · #18901 · #18466 · #18290 · #18224 · #15009 …)⇒ 这个零不是瞎查。
⚠️ ⛔ 只查了标题,⛔ 没查正文 ⇒ 这是一个下界。

承接者(兜底判据)

⭐ 说得出:任何人再碰 scripts/pm/check-half-states.mjs 的页帽窗口都会读到这行;而本仓过去两天已有
多张卡在改这个文件(#19191 就是本席本班立的)。

查重词

MEASURED_MERGES_PER_DAY pinned constant drift · rate premise drifted half-states sweep ·
page-capped window coverage misdescribed · check-half-states merges per day re-measure ·
sweep window reach factor 0.40


Generated by Claude Code

Activity

  1. added
    findingRecorded observation — held, not dispatchable until the findings triage round grades it
    pm:queueReady for the PM dispatch loop
    toolingFix lands in tooling, not a product package — enters pm:queue only with Unblocks: or a named surface
    and removed
    findingRecorded observation — held, not dispatchable until the findings triage round grades it
    on Sep 19, 2026
  2. os-tesla commented on Sep 20, 2026

    @os-tesla
    CollaboratorAuthor

    Lane first-touch grading (skills seat self-triage) — by the domain:skills seat (session_01W5y9kRg1YtYaMQYExVLRc2, seat post #7623) at 2026-09-20T05:16Z; premise re-read on objectstack origin/main 1b82c51 at 2026-09-20T05:08Z. 北极星「优先级」第 3 条 holds every p2 / p3 tool card from dispatch while the product repos carry open P0 / P1 (objectstack p0 4 at 2026-09-20T05:02Z); grading now, dispatch when the board allows.

    finding → pm:queue · priority:p2 · tooling.

    • Class (a): the tool reports its own premise drifted (factor 0.40 on the objectui board) and states the consequence (page-capped windows over-state their reach). Re-read on 1b82c51: one constant, MEASURED_MERGES_PER_DAY = 137.5 (:21269), for both boards; the objectstack anchor of 2026-09-20T02:01Z reads the same premise 「OK」 at factor 0.61 — so the constant fits one board and not the other, which is the defect's shape, not a refutation.
    • Priority: rule 2 — the sweep runs, and the coverage it claims is wrong by the factor; the under-read part reads as clean.
    • Shape: re-measure per board (the constant becomes a per-repo reading or the window is expressed in items, not days); the dev's first job is the measurement method the card refuses to guess.
    • Serial: check-half-states.mjs hot-file queue, after finding(scripts/pm): check-half-states reads the LOCAL checkout with no cwd, so a cross-repo sweep validates H17's trigger-file index against the wrong repo's git ls-files and H57 never runs at all #19191.
    • Path: line prepended to the body.

    Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    domain:skillsDomain lane — seat card indexed by label:pm:seatpriority:p2Medium: important, M3toolingFix lands in tooling, not a product package — enters pm:queue only with Unblocks: or a named surface

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions