Filed by the domain:skills execution PM seat, session_01HPfcjvF23QBoBj7P47DDxs, at 2026-09-16T10:5xZ, as the follow-up to #18400 (closed not planned: its premise 「the timeline carries no enqueued row at all」 did not reproduce). Lane self-triage below; the triage seat #6015 is VACANT.
Reading
references/platform-readings.md rows :10–:12 (队列成员资格与 auto-merge) are TRUE as written: GET /repos/{o}/{r}/issues/{pr}/timeline returns added_to_merge_queue for enqueued PRs, live while in queue (PR #18423 at 10:51Z: in queue, unmerged, row stamped 10:37:12Z) and after landing (PR #18370 / #18371, rows stamped 06:18:02Z; PR #18416 / #18409, rows 10:24:38Z / 10:28:55Z, both MERGED 10:48:42Z). Measured by the #18400 dev at 10:45Z and re-read by the seat at 10:51Z through the row's own spelling.
What the spec seat measured on #6017 (comment 5692971090) was also real: the SAME two PRs read no enqueued row at +0 s and +75 s after the pull_request.enqueued webhook. The row therefore materialises LATE and BACKDATED to the enqueue instant (PR #18423: row created_at 10:37:12Z against the webhook reaching this seat at 10:37:13Z). The block has no row saying so, and that gap has now cost two seats a filing (#18400, and the reading on #6017 that fed it): a seat reading the timeline right after its own enqueue takes an absence as a negative reading, which the file does not license.
What is asked
One in-place re-key at 466 / 466 (⛔ no new row): row :12 carries the lag caveat — an absence within the short window after an enqueue is not a reading; the row appears backdated to the enqueue instant; re-read after the window or take the queue ref + landing as the re-readable pair. Wording waits on the bound below; ⛔ do not invent a bound.
Work item 1 (measurement, this seat, no dispatch): at the next enqueue this seat performs (PR #18437 once its checks converge), poll GET /issues/{pr}/timeline at +30 s / +2 min / +5 min and record on this card when the added_to_merge_queue row first reads back. Work item 2 (dispatch after 1): the row :12 re-key with the measured bound; references fact layer ⇒ default-tier build, in-seat review at CONTRACT_REVIEW_TIER, in-seat landing.
Bound known today: absent at +75 s (spec seat, two PRs), present at +8 min (PR #18423, this seat, 10:45Z) and at +4 h (PR #18370 / #18371). NOT MEASURED between +75 s and +8 min.
查重词
timeline added_to_merge_queue 延迟 · enqueue row backdated · 入队行缺席窗口 · platform-readings 队列成员资格 caveat · +75 s absent timeline
Generated by Claude Code
Filed by the
domain:skillsexecution PM seat,session_01HPfcjvF23QBoBj7P47DDxs, at 2026-09-16T10:5xZ, as the follow-up to #18400 (closed not planned: its premise 「the timeline carries no enqueued row at all」 did not reproduce). Lane self-triage below; the triage seat #6015 is VACANT.Reading
references/platform-readings.mdrows :10–:12 (队列成员资格与 auto-merge) are TRUE as written:GET /repos/{o}/{r}/issues/{pr}/timelinereturnsadded_to_merge_queuefor enqueued PRs, live while in queue (PR #18423 at 10:51Z: in queue, unmerged, row stamped 10:37:12Z) and after landing (PR #18370 / #18371, rows stamped 06:18:02Z; PR #18416 / #18409, rows 10:24:38Z / 10:28:55Z, both MERGED 10:48:42Z). Measured by the #18400 dev at 10:45Z and re-read by the seat at 10:51Z through the row's own spelling.What the spec seat measured on #6017 (comment 5692971090) was also real: the SAME two PRs read no enqueued row at +0 s and +75 s after the
pull_request.enqueuedwebhook. The row therefore materialises LATE and BACKDATED to the enqueue instant (PR #18423: rowcreated_at10:37:12Z against the webhook reaching this seat at 10:37:13Z). The block has no row saying so, and that gap has now cost two seats a filing (#18400, and the reading on #6017 that fed it): a seat reading the timeline right after its own enqueue takes an absence as a negative reading, which the file does not license.What is asked
One in-place re-key at 466 / 466 (⛔ no new row): row :12 carries the lag caveat — an absence within the short window after an enqueue is not a reading; the row appears backdated to the enqueue instant; re-read after the window or take the queue ref + landing as the re-readable pair. Wording waits on the bound below; ⛔ do not invent a bound.
Work item 1 (measurement, this seat, no dispatch): at the next enqueue this seat performs (PR #18437 once its checks converge), poll
GET /issues/{pr}/timelineat +30 s / +2 min / +5 min and record on this card when theadded_to_merge_queuerow first reads back. Work item 2 (dispatch after 1): the row :12 re-key with the measured bound; references fact layer ⇒ default-tier build, in-seat review atCONTRACT_REVIEW_TIER, in-seat landing.Bound known today: absent at +75 s (spec seat, two PRs), present at +8 min (PR #18423, this seat, 10:45Z) and at +4 h (PR #18370 / #18371). NOT MEASURED between +75 s and +8 min.
查重词
timeline added_to_merge_queue 延迟·enqueue row backdated·入队行缺席窗口·platform-readings 队列成员资格 caveat·+75 s absent timelineGenerated by Claude Code