perf(webhook): give every coalesce family the quiet window its key needs

[JSONbored]

Closes #10127

githubWebhookCoalesceKey computes a job coalesce key for five event families. githubWebhookCoalesceDelaySeconds returned a non-zero window for exactly one of them — the push family (#9479) — and 0 for the rest. A zero-delay job is claimed off the queue before a sibling can arrive, so there is no pending row to merge into.

The file already says this, in the push section:

On a queue that coalesces (self-host pg/sqlite, which is where the ORB runs) this is what CREATES the window: without a delay the first push is claimed immediately and later pushes find no pending row to merge into.

That reasoning was never extended to the other four. Their keys were derived, attached to the job, and structurally unable to do anything.

Measured on the Orb, not projected

 event_name  | action    | deliveries (since 2026-07-30)
 check_suite | completed |       2143

 records | heads | per_head
    1528 |   519 |    2.944

Nearly three full gate passes per distinct head SHA. Worked example — #10098, eleven hold records at one head SHA in 33 minutes, every one upstream_state_change.

The LLM review is not duplicated. ai_review_pr shows 415 calls against 519 distinct heads, so the per-head review cache is holding. The waste is GitHub REST/GraphQL quota, queue churn, decision_records writes, and label/comment reconciliation — which is exactly why it went unnoticed: the cost line looked fine.

The change

The key and its window now come from one classifier, so a family cannot have one without the other:

type WebhookCoalescePlan = { key: string; quietWindowSeconds: number };
function planGithubWebhookCoalesce(eventName, payload): WebhookCoalescePlan | null;

Both exported functions become projections of it. That coupling is the actual fix — the numbers below are tuning.

family	window	why
pr-push	45s	unchanged (#9479)
ci-completed	45s	interchangeable — whichever delivery wins re-fetches the same settled CI state. Kept under CI_COALESCE_WINDOW_SECONDS (60s) so the downstream per-PR throttle stays the outer bound
pr-label	15s	shorter: a maintainer removing a hold label expects prompt notice
review-surface	15s	payload-interchangeable within a family
pr-refresh	0	lifecycle — opened/reopened/ready_for_review each happen once and start the clock a contributor waits on. Keyed so a same-instant duplicate still collapses, never delayed

Lifecycle keeping a zero window is deliberate. Delaying a PR's first verdict to save a redelivery trades a real cost for an imaginary one: 203 opened deliveries against 1,727 CI completions in the same period — lifecycle is not a burst source.

Verification

• Full suite and the whole check-* sweep green.
• The invariant is asserted as "key implies window" over the real families rather than against expected constants, so a family added later cannot quietly reintroduce a keyed-but-inert delivery. That framing also survived the constants becoming module-local when dead-exports:check (correctly) objected to exporting them.
• Mutation-tested:

mutation	tests failed
CI window → 0 (literally the original bug)	1
surface window → 0	1
lifecycle gains a 30s delay (the opposite error)	1

What I am not claiming

How much this actually collapses depends on how CI suites cluster in time, which I cannot predict from here. Baseline is captured (2.944 records/head at 2026-07-31T08:14:58Z); I will re-measure on the Orb after the deploy and report the real figure rather than a projection.

Separately checked and found already handled, so deliberately not touched: the 2,064 issue_comment deliveries are the bot's own comment churn, and isSelfAuthoredAppCommentWebhook / isBotAuthoredIssueCommentEditWebhook already drop them at ingest before any gate pass.

[loopover-orb]

Important

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏳ LoopOver is waiting…

LoopOver has seen this pull request and is waiting on CI checks to finish before reviewing it. This comment will update once the review runs.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed · 🟨 Waiting}

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.09%. Comparing base (645b323) to head (37bd684).
⚠️ Report is 7 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #10133      +/-   ##
==========================================
- Coverage   91.96%   91.09%   -0.88%     
==========================================
  Files         931      931              
  Lines      113966   113972       +6     
  Branches    27513    27514       +1     
==========================================
- Hits       104808   103820     -988     
- Misses       7858     9047    +1189     
+ Partials     1300     1105     -195     

Flag	Coverage Δ
backend	94.12% <100.00%> (-1.56%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/github/webhook-coalesce.ts	100.00% <100.00%> (ø)

... and 3 files with indirect coverage changes