[Drift Audit] Correct interval API and Duration type in the wide-events rule doc

[daxmobile]

Task/Issue URL: (app.asana.com/redacted)

Description

.cursor/rules/wide-events.mdc — the interval section and the CleanupPolicy.OnTimeout example.

The wide-events area changed in this window via #7965 — Add support for wide event sampling (2ac7b09149). That PR updated the doc for sampling itself, and I verified the new Sampling section against WideEventClientImpl — SAMPLED_OUT_FLOW_ID = -1L, the if (wideEventId == SAMPLED_OUT_FLOW_ID) return Result.success(...) short-circuit in every method, and the persisted probability — it is accurate and I left it alone. Re-reading the rest of the doc against the API surfaced two claims that no longer hold.

1. intervalStart — described as a two-argument call with hand-bucketed keys.

What the doc described: intervalStart(wideEventId, key) under the comment "always use bucketed values, not raw ms", which reads as an instruction to bucket at the call site via the _ms_bucketed naming convention.

The code change that invalidated it: #8381 — Configurable interval buckets on WideEventClient.intervalStart (a8076fbe81). intervalStart now takes timeout: Duration? = null and buckets: Set<Duration>? = DEFAULT_INTERVAL_BUCKETS, and the client does the bucketing — rounding down to the nearest boundary, recording 0 below the smallest, and recording the raw duration when buckets is null. Real callers pass custom bucket sets (PageLoadWideEvent, PirScanWideEvent) and timeouts (SubscriptionSwitchWideEvent at 10.minutes, SubscriptionPurchaseWideEvent at 4.hours) — none of which the doc mentioned.

The timeout gap also left a dangling concept: the CleanupPolicy section documents ignoreIfIntervalTimeoutPresent as "keep alive if an interval timeout is set", but nothing in the doc explained what sets one.

How it now reads: the API example shows intervalStart with timeout and buckets, and a new Intervals subsection states the rounding-down behaviour, the default bucket set, what null does, and ties the timeout parameter back to ignoreIfIntervalTimeoutPresent.

2. CleanupPolicy.OnTimeout(duration = Duration.ofDays(7)) no longer compiles.

The same PR migrated the API from java.time.Duration to kotlin.time.Duration — flowStart's default is now OnTimeout(duration = 7.days) and intervalEnd returns Result<kotlin.time.Duration>. The doc's Duration.ofDays(7) is a java.time call against a kotlin.time parameter. Now reads 7.days, with a line naming the type so the 10.minutes / 7.days literals aren't guesswork.

I also checked, and did not treat as drift:

• .cursor/rules/pixels.mdc / pixel-definitions.mdc vs #9306 (unified_input_surface added to unified-input action pixels, plus a params_dictionary entry) and the PIR pixel additions in PirPixel/PirPixelSender/PirPixelInterceptor — routine additions that follow the documented shape; no documented claim became false.
• .cursor/rules/android-design-system.mdc — Theming.kt and MessageCta.kt saw small edits, but nothing the doc describes. Note this doc already has two open Drift Audit PRs against it (#9300, #9315); I stayed out of it to avoid conflicting edits.
• .cursor/rules/maestro-ui-tests.mdc — three .maestro/** files changed by one line each (onboarding test content), not the tags/config/run mechanics the doc describes.
• .cursor/rules/architecture.mdc — the remaining merged PRs in the window (simplified sync UI, native input, onboarding dialogs, content-scope-scripts bumps, PIR renderer-crash handling, Glide cache fix, Develocity plugin bump) are all inside -impl code or build config the rule docs don't describe.

One thing for a human to weigh: the interval drift dates from #8381 (30 Apr), not from this window's #7965. I fixed it because re-checking the wide-events doc is exactly what a change in that area calls for and the mismatch is unambiguous — the example doesn't compile — but if the preference is to keep each Drift Audit PR strictly to its window, that is a fair reason to trim this one.

.cursor/rules/wide-events.mdc is the only file changed — no code, and the .claude/rules symlink is untouched.

🤖 AI-docs Drift Auditor

Steps to test this PR

• Read the updated Intervals section against WideEventClient.intervalStart's KDoc and DEFAULT_INTERVAL_BUCKETS in statistics/statistics-api/.../wideevents/WideEventClient.kt, and confirm the rounding-down / zero-below-smallest / null-disables-bucketing description matches.
• Confirm the Duration type claim: WideEventClient.kt imports kotlin.time.Duration and defaults flowStart's cleanup policy to OnTimeout(duration = 7.days), so the corrected 7.days example compiles and Duration.ofDays(7) would not.
• Spot-check the named callers — PageLoadWideEvent / PirScanWideEvent pass custom buckets, SubscriptionSwitchWideEvent passes timeout = 10.minutes — and confirm the doc's timeout / buckets guidance reflects real usage.

UI changes

Before	After
No UI changes	No UI changes

Note

🔒 Integrity filter blocked 4 items

The following items were blocked because they don't meet the GitHub integrity level.

• #8891 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #8809 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #8811 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #9341 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by AI-docs Semantic Drift Audit · ◷